Académique Documents
Professionnel Documents
Culture Documents
Research - Meaning - Types - Nature and scope of research - Problem formulation statement of research objective - value and cost of information - Decision theory Organizational structure of research - Research process - research designs - exploratory
-Descriptive - Experimental research.
OBJECTIVE
To equip the students with the basic understanding of the research methodology and
provide an insight into the application of modern analytical tools and techniques for the
purpose of management decision making.
STRUCTURE
LEARNING OBJECTIVES
INTRODUCTION
The task of business research is to generate accurate information for use in decision
making. The emphasis of business research is shifting the decision makers from
intuitive information that is based on own judgment and gathering information into
systematic and objective investigation.
DEFINITION
The business research is defined as the systematic and objective process of gathering,
recording and analyzing data for aid in making business decisions. Literally, research
means to "search again". It connotes patient study and scientific investigation where in
the research takes more careful look to discover to know about the subject of study. The
data collected and analyzed are to be accurate, the research need to be very objective.
Thus, the role of researcher is to be detached and impersonal rather than engaging in
biased attempt. This means without objectivity the research is useless. The definition is
restricted to take decision in the aspects of business alone. This generates and provides
the necessary qualitative and quantitative information upon which, as base, the
decisions are taken. This information reducing the uncertainly of decisions and reduces
the risk of making wrong decisions. However research should be an "aid" to managerial
judgment, not a substitute for it. There is more to management than research. Applying
research remains a managerial art.
The study of research methods provides with the knowledge and skills that need to solve
the problems and meet the challenges of the fast-paced decision making environment.
There are two important factors stimulate an interest in scientific approach to decision
making:
1. The is an increased need for more and better information, and
2. The availability of technical tools to meet this need.
During the last decade, we have witnessed dramatic changes in the business
environment. These changes have created new knowledge needs for the manager to
consider when evaluating any decision. The trend toward complexity has increased the
risk associated with business decision, making it more important to have sound
information base. The following are the few reasons which makes the researcher to
lookout for newer and better information based on which the decisions are taken:
The development of scientific method in business research lags behind the similar
developments in physical science research which is more rigorous and much more
advanced. But business research is of recent origin and moreover the finding cannot be
patented that of physical science research. Business research normally deals with topics
such as human attitudes, behavior and performance. Even with these hindrances,
business research is making strides in the scientific arena. Hence, the managers who are
not proposed ior this scientific application in business research will be at severe
disadvantage.
Value of Business Research
3. Nature of Information
The value of research will depend upon the nature of decisions to be made. A routine
decision does not require substantial information or warrants. However, for important
and strategic decision, more likely research needs to be conducted.
4. Benefits vs. Costs
The decision to conduct research boils down to these important questions.
1. Will the rate of return be worth the investment?
2. Will the information improve the quality of the decision?
3. Is the research expenditure the best use of available funds?
Thus, the cost of information should not exceed the benefits i.e. value of information.
What is Good Research?
Good research generates dependable data can be used reliable for making managerial
decisions. The following are the tips of good research.
Scope of Research
The scope of research on management is limited to business. A researcher conducting
research within an organization may be referred as a "marketing researcher" or
"organizational researcher", although business research is specialized and the term
encompasses all the functional areas Production, Finance, Marketing, HR etc. The
different functional areas may investigate different phenomenon, but they are
comparable to one another because they use similar research methods. There are many
kinds of areas are resembled in the business environment like forecasting, trends
environment, capital formation, portfolio analysis, cost analysis, risk analysis, TQM, job
satisfaction, organizational effectiveness, climate, culture, market potential,
segmentation, sales analysis, distribution channel, computer information needs
analysis, social values and establish and etc.
Types of Research
Research is to develop and evaluate concepts and theories. In broader sense research
can be classified as.
1) Basic Research or Pure Research
It does not directly involve the solution to a particular problem. Although basic research
generally cannot be implemented, this is conducted to verify the acceptability of a given
theory or to discuss more about a certain concept.
2) Applied Research
It is conducted when a decision must be made about a specific real-life problem. It
encompasses those studies undertaken to answer question to specific problems or to
make decision about particular course of action.
However, the procedures and techniques utilized by both researchers do not differ
substantially. Both employ scientific method to answer questions. Broadly, the scientific
method refers to techniques and procedures that help researcher to know and
understand business phenomenon. The scientific method requires systematic analysis
and logical interpretation of empirical evidence (facts from observation or
experimentation) to confirm or dispose prior conceptions. In basic research, it first tests
the prior conceptions or assumptions or hypothesis and then makes inferences and
conclusions. In the applied research the use of scientific method assures objectivity in
gathering facts and taking decision.
At the outset, it may be noted that there are several ways of studying and tackling a
problem. There is no single perfect design. Research designs have been classified by
authors in different ways. Different types of research designs have emerged on the
account of the different perspectives from which the problem or opportunity is viewed.
However, the research designs broadly classified into three categories - exploratory,
descriptive, and causal research. The research can be classified on the basis of either
technique or function. Experiment, surveys and observation are few common
techniques. The technique may be qualitative of quantitative. Based on the nature of the
problems or purpose of study the above three are used invariably used in management
parlance.
3) Exploratory Research
The focus is mainly on discovering of ideas. An exploratory research is generally based
on secondary data that are already available. It is to be understood that this type of
study is conducted to classify ambiguous problems. These studies provide information
to use in analyzing situations. This will helps to crystallize a problem and identify
information needs for further research. The purpose of exploration is usually to develop
hypotheses or question for further research. The exploration may be accomplished with
different techniques. Both qualitative and quantitative techniques are applicable
is unsure of what is to be done and ii) where extreme profits or losses involved. A
pertinent question is-how much information should be collected in a given situation?
Since the collected information involves a cost, it is necessary to ensure that the benefit
from the information is more than the cost involved in its collection.
Decision Theory
With reference to the above discussion, an attempt is needed to see how information can
be evaluated for setting up a limit. The concept of probability is the basis of decision
maker under conditions of uncertainly. These are three basic sources of assigning
probabilities
1) Based on a logic / deduction: For e.g., when a coin is tossed, the probability of
getting a head or tail is 0.5.
2) Past experience / Empirical evidence: The experience gained in the process
resolving these problems in the past. On the basis of its past experience, it may be in a
better positive to estimate the probability of new decisions.
3) Subjective Estimate: The most frequently used method, it is based on the
knowledge and information with respect to researcher for the probability estimates.
The above discussion was confined to single stage problem wherein the researcher is
required to select the best course of action on the basis of information available at a
point at time. However, there are problems with multiple stages wherein a sequence of
decisions involved. Each decision leads to a chance event which in turn influences the
next decision. In those cases, a Decision Tree Analysis i.e. graphical derives depicting
the sequence of action-event combination, will be useful in making a choice between two
alternatives. If the decision tree is not helpful, more sophisticated technological known
as Bayesian Analysis can be used. Here, the probabilities can be revised on account of
the availability of new information using prior, posterior and pre-posterior analysis.
There is a great deal between budgeting and value assessment in management decision
to conduct research. An appropriate research study should help managers avoid losses
and increase sales or profits; otherwise, research can be wasteful. The decision maker
wants a cost-estimate for a research project and equally precise assurance that useful
information will result from the research. Even if the researcher can give good cost and
information estimates, the decision maker or manager still must judge whether the
benefits out-weigh the costs.
Conceptually, the value of research information is not difficult to determine. In business
situation the research should provide added revenues or reduce expenses. The value of
research information may be judge in terms of the difference between the results of
decisions made with the information and the result that would be made without it. It is
simple to state, in actual application, it presents difficult measurement problems.
Guideline for approximately the cost-to-value of Research
1. Focus on the most important issues of the project: Identify certain issues as
important and others as peripheral to the problems. Unimportant issues are only to
drain resources.
2. Never try to do much: There is limit to the amount of information that can be
collected. The researcher must take a trade-off between the number of issues that can be
dealt with the depth of each issue. Therefore it is necessary to focus on those issues of
greatest potential value.
3. Determine whether secondary, primary information or combination is
needed: The most appropriate must be selected that should address the stated
problem.
4. Analyze all potential methods of collecting information: Alternative data
sources and research designs are available that will allow detailed investigation of issues
at a relatively low cost.
5. Subjectively asses the value of information: The researcher need to ask some
fundamental questions relating to objections. For example, a) Can the information be
collected at all? b) Can the information tell something more that already what we have?
c) Will the information provide significant insights? d) What benefits will be delivered
from this information?
Structure of Research
Business research can take many forms, but systematic inquiry is a common thread.
Systematic inquiry requires an orderly investigation. Business research is a sequence of
highly interrelated activity. The steps research process overlap continuously.
Nevertheless, research on management often follows a general pattern. The styles are:
1. Defining the problem
2. Planning a research design
3. Planning a sample
4. Collecting data
5. Analyzing the data
6. Formulating the conclusions and preparing the report
SUMMARY
This paper outlined the importance of business research. Difference between basic and
applied research have been dealt in detail. This chapter has given the meaning, scope,
types, and structure of the research.
KEY TERMS
Research
Value of research
Need for research
Good Research
Scope of research
Types of research
Basic research
Applied research
Scientific method
Exploratory research
Descriptive research
Cross - sectional and longitudinal
Causal research
Decision theory
Structure
QUESTIONS
1. What are some examples of business research in your particular field of interest?
2. What do you mean by research? Explain its significance in modem times.
3. What is the difference between applied and basic research?
4. What is good research?
5. Discuss: Explorative, descriptive and causal research.
6. Discuss the value of information and cost using decision theory.
7. Discuss the structure of business research.
OBJECTIVES
To discuss the nature of decision makers objectives and the role they play in
defining the research
To understand that proper problem definition is essential for effective business
research
To discuss the influence of the statement of the business problem on the specific
research objectives
To state research problem in terms of clear and precise research objectives.
STRUCTURE
Problem definition
Situation analysis
Measurable Symptoms
Unit of analysis
Hypothesis and Research objectives
PROBLEM DEFINITION
Before choosing a research design, manager and researcher need a sense of direction for
the investigation. It is extremely important to define the business problem carefully
because the definition determines the purposes of the research and ultimately the
research design. Well defined problem is half solved problems - hence, researcher must
understand how to define problem. The formal quantitative research process should not
begin until the problem has been clearly defined. Determination of research problem
consists of three important tasks namely,
1. Classifying in argument information needs.
2. Redefining research problem, and
3. Establishing hypothesis & research objectives.
Step 1: To ensure that appropriate information created through this process, researcher
must assist decision maker in making sure that the problem or opportunity has been
clearly defined and the decision maker is aware of the information requirements. This
include the following activities namely,
i) Purpose: Here, the decision maker holds the responsibility of addressing a
recognized decision problem or opportunity. The researcher begins the process by
asking the decision maker to express his or her reasons for thinking there is a need to
undertake research. By this questioning process, the researcher can develop insights as
what they believe to be the problems. One method that might be employed to familiarize
the decision maker is the iceberg principle. The dangerous part of many problems,
like submersed portion of the iceberg, is neither visible nor understood by managers. If
the submerged position of the problem is omitted from the problem definition, then
result may be less than optimal.
ii) Situation Analysis or understanding the situation: To gain the complete
understanding, both should perform a basic situation analysis of the circumstances
surrounding the problem area. A situational analysis is a popular tool that focuses on
the informal gathering of background information to familiarize the overall complexity
of the decision. A situation analysis attempts to identify the event and factors that have
led to the current decision problem situation. To objectively understand the client's
domain (i.e., industry, competition, product line, markets etc) the researchers not rely
only on the information provided by client but also others. In short the researcher must
develop expertise in the client's business.
iii) Identifying and separating measurable symptoms: Once the researcher
understands the overall problem situation, they must work with decision maker to
separate the problems from the observable and measurable symptoms that may have
been initially perceived as the being the decision problem.
iv) Determining unit of analysis: The researcher must be able to specify whether
data should be collected about individual, households, organizations, departments,
geographical areas, specific object or some contributions of these. The unit of analysis
will provide direction in later activities such as scale measurement development and
drawing appropriate sample of respondents.
v) Determining relevant variables: Here, the focus is on the identifying the
different independent or dependent variables. It is determination of type of information
(i.e., facts, estimates, predictions, relationships) and specific constructs (i.e. concepts or
ideas about an object, attributes, or phenomenon that are worth measurement)
Step 2:
Once the problem is understood and specific information requirements are identified,
then the researcher must redefine the problem in more specific terms. In reframing the
problems and questions as information research questions, they must use their scientific
knowledge expertise. Establishing research questions specific to problems will force the
decision maker to provide additional information that is relevant to the actual problems.
In other situations, redefining problems as research problems can lead to the
establishment of research hypothesis rather than questions.
Step 3: (Hypothesis & Research objective)
A hypothesis is basically an unproven started of a research question in a testable
format. Hypothetical statement can be formulated about any variable and can express a
possible relationship between two or more variables. While research questions and
hypotheses are similar in their intent to express relationship, the hypotheses tend to be
more specific and declarative, whereas research questions are more interrogative. In
other words hypotheses are statement that can be empirically tested.
Research objectives are precise statements of what a research project will attempt to
achieve. It indirectly represents a blueprint of research activities. Research objectives
allow the researcher to document concise, measurable and realistic events that either
increase or decrease the magnitude of management problems. More importantly it
Problem definition
Iceberg principle
Situation analysis
Unit of analysis
Variables
Hypotheses
Research objectives
QUESTIONS
1. What is the task of problem definition?
2. What is the iceberg principle?
3. State a problem in your field of interest, and list some variables that might be
investigated to solve this problem.
4. What do you mean by hypothesis?
5. What is a research objective?
OBJECTIVES
STRUCTURE
Research process
Research design
Types of research designs
Explorative research
Descriptive research
Casual research
Research Process
Before discussing the phases and specific steps of research process, it is important to
emphasize the need for information and when the research is conducted or not. In this
context, the research process may be called as information research process that would
be more appropriate in business parlance. The information research is used to reflect
the evolving changes occurring within the management and the rapid changes facing
many decision makers regarding how firms conduct both internal and external
activities. Hence, understanding the process of transforming raw data into usable
information from broader information and expanding the applicability of the research
process in solving business problems and opportunities is very important.
Overview: The research process has been described anywhere from 6 to 11
standardized stages. Here, the process consist of four distinct inter related phases that
have logical, hierarchical ordering depicted below.
Diagram: Four phases of Research Process
The following exhibit represents the interrelated steps of the 4 phases of the research
process. Generally researchers should follow the steps in order. However, the
complexity of the problem, the level of risk involved and management needs will
determine the order of the process.
Exhibit
Phase 1: Determination of Research Problem
Step 1: Determining management information needs
Step 2: Redefining the decision problem as research problem.
Step 3: Establishing research objectives.
Phase 2: Development of Research Design
Step 4: Determining to evaluate research design.
Step 5: Determining the data source.
Step 6: Determining the sample plan and sample size.
Step 7: Determining the measurement scales.
Phase 3: Execution of the Research Design
Step 8: Data collection and processing data.
Step 9: Analysing the data.
Phase 4: Communication of the Results
Step 10: Preparing and presenting the final report to management.
Step 1: Determining management information needs
Before the researcher becomes involved usually, the decision maker has to make a
formal statement of what they believe is the issue. At this point, the researcher's
responsibility is to make sure management has clearly and correctly specified the
opportunity or question. It is important for the decision maker and the researcher to
agree on the definition of the problem so that the result of the research process will
produce useful information. Actually the researcher should assist the decision maker in
determining whether the referred problem is really a problem or just a symptom or a yet
unidentified problem. Finally the researchers list the factors that could have a direct or
indirect impact on the defined problem or opportunity.
choose between a sample (small population) and census (entire population). To achieve
this research objective, the researcher needs to develop explicit sampling plan which will
serve as a blueprint for defining the target population. Sampling plans can be classified
into two general types: probability (equal chance) and non probability. Since sampling
size affects quality and general ability, researchers, must think carefully about how
many people to include or how many objects to investigate.
Step 7: Determining the measurement scales.
This step focuses on determining the dimensions of the factors being investigated and
measuring the variables that underlie the defined problem. This determines how much
raw data can be collected and the amount of data to be collected. The level of
information (nominal, ordinal, interval, and ratios), the reliability, the validity and
dimension (uni vs. multi) will determine the measurement process.
Step 8: Data collection and processing data.
There are two fundamental approaches to gather raw data. One is to ask questions about
variables and phenomena using trained interviewers or questionnaires. The other is to
observe variables or phenomena using professional observers or high tech mechanical
devices. Self - administered surveys, personal interviews, computer simulations,
telephone interviews are some of the tools to collect data. The questioning allows a
wider variety of collective of data about not only past, present but also the state of mind
or intentions. Observation can be characterized as natural, contrived, disguised or
undisguised, structured or unstructured, direct or indirect, human or mechanical, and
uses the devices like video camera, tape recorders, audiometer, eye camera, psychogalvanometer or pupil meter. After the raw data collected, a coding scheme is needed so
that the raw data can be entered into computers. It is assigning logical numerical
description to all response categories. The researcher must then clean the raw data of
either coding or data entry error.
Step 9: Analysing the data.
Using a variety of data analysis technique, the researcher can create new, complex data
structure by continuing two or more variables into indexes, ratios, constructs and so on.
Analysis can vary from simple frequency distribution (percentage) to sample statistic
measures (mode, median, mean, standard deviation, and standard error) to multivariate
data analysis.
Step 10: Preparing and presenting the final report to management.
This step is to prepare and present the final research report to management. The report
should contain executive summary, introduction, problem definition and objectives,
methodology, analysis, results and finding, finally suggestions and recommendation. It
also includes appendix. Any researcher is expected not only submit well produced
written report but also oral presentation.
Research Design
Kerlinger, in his "Foundations of Behavioral Research" book defines, "Research design
is the plan structure, and strategy of investigation conceived so as to obtain answers to
research questions and to control variance". The plan is overall scheme or program of
the research. It includes an outline of what the investigator will do from writing
hypotheses and their operational implication to the final analysis of data. A structure is
the framework, and the relations among variables of a study. According to Green & Tull,
a research design is the specification of methods and procedures for acquiring the
information needed. It is the overall operational pattern or framework of the project
that stipulates what information is to be collected from which sources by what
procedures.
From the above definitions it can be understood that the research design is more or less
a blueprint of research, which lays down the methods and procedure for requisite
collection of information and measurement and analysis with a view to arrive at
meaningful conclusions of the proposed study.
Types of Research Design
The different types of design are explained in the previous section (refer types of
research). There are three frequently used classification is give below.
I. Explorative
II. Descriptive
III. Casual
Here the focus will be how these studies are conducted and methods are explained:
I. Categories of Explorative Research
There are four general categories of explorative research methods. Each category
provides various alternative ways of getting information.
1. Experience Surveys
It is an attempt to discuss issues and ideas with top executive and knowledge
people who have experience in the field. This research in the form of experience
survey may be quite informal. This activity intends only to get ideas about the'
problems. Often an experience survey may consist of interviews with a small
number of people who have been carefully selected. The respondents will
generally be allowed to discuss the questions with few constraints. Hence, the
purpose of such experts is to help formulate the problem and classify concepts
rather than develop conclusive evidence.
O1
O2
X
X
O1
O2
O1
O3
X
X
O2
O4
As the diagram above indicated, the subjects of the experimental group are tested
before and after being exposed to the treatment. The control group is tested twice,
at the same time of experimental group, but these subjects are not exposed to the
treatment. The effect is calculated as follows:
(O2 - O1) - (O4 O3)
5. Four-group, six- study design: (Solomon four groups design) Combining,
the before - after with control group design and the after-only with control group
design provides a means for controlling testing affects, as well as other sources of
extraneous variations. The diagram as follows:
Experimental Group 1
Control Group 1
O1
O3
Experimental Group 2:
O4
Control Group 2:
O2
O5
O6
6. Time series design: When experiments are conducted over long periods of
time, they are more vulnerable to history effects due to changes in population,
attitudes, economic patterns and the like. Hence, this is also called quasiexperimental design. This design can be diagrammed as follows:
O1
O2
O3
O4
O5
O6
Several observations have been taken before and after the treatment to determine
the patterns after the treatment are similar to the pattern of before the treatments.
7. Completely randomized design (CRD): CRD is an experimental design
that uses a random process to assign experimental units to treatments. Here,
randomization of experimental units to control extraneous variables while
manipulating a single independent variable, the treatment variable.
8. Randomized block design (RBD): The RBD is an extension of the CRD. A
form of randomization is utilized to control for most extraneous variation. Here an
attempt is made to isolate the effects of the single variable by blocking its effects.
9. Factorial design: A FD allows for testing the effects of two or more
treatment (factors) at various levels. It allows for the simultaneous manipulation
of 2 or more variable at various levels. This design will measure main effect (i.e.,
the influence on the dependent variable by each independent variable and also
interaction effect.
10. Latin Square Design (LSD): The LSD attempts to control or block out the
effect of two or more confounding extraneous factors. This design is so named
because of the layout of the table that represents the design.
SUMMARY
This chapter has outlined the stages in business research prawn. Various types of
research design have been dealt in detail.
KEY TERMS
Exploratory research
Descriptive research
Causal research
Focus group
Projective techniques
Case studies
Depth Interview
Experimental designs
Quasi experimental design
True experimental designs
Complex designs
QUESTIONS
1. Explain the different phases of research process.
2. Briefly describe the different steps involved in a research process.
3. What are major types researches in business?
4. Discuss the categories of exploratory and descriptive research.
5. Explain different experimental designs.
REFERENCES
1. Bellenger and et al, Marketing Research, Home Wood Illinois, Inc. 1978.
2. Boot, John C.G. and Cox., Edwin B., Statistical Analysis for Managerial Decisions, 2nd
ed. New Delhi: McGraw Hill Publishing Co. Ltd.
3. Edwards, Allen, Statistical Methods, 2nd ed., New York. 1967.
- End of Chapter -
LESSON 4
INTRODUCTION TO STATISTICS AND POPULATION PARAMETERS
OBJECTIVES
STRUCTURE
INTRODUCTION
At the outset, it may be noted that the word 'Statistics' is used rather curiously in two
senses-plural and singular. In the plural sense, it refers to a set of figures. Thus, we
speak of production and sale of textiles, television sets, and so on. In the singular sense,
Statistics refers to the whole body of analytical tools that are used to collect the figures,
organize and interpret them and, finally, to draw conclusions from them.
It should be noted that both the aspects of Statistics are important if the quantitative
data are to serve their purpose. If Statistics, as a subject, is inadequate and consists of
poor methodology, we would not know the right procedure to extract from the data the
information they contain. On the other hand, if our figures are defective in the sense
that they are inadequate or inaccurate, we would not reach the right conclusions even
though our subject is well developed. With this brief introduction, let us first see how
Statistics has been defined.
Statistics has been defined by various authors differently. In the initial period the role of
Statistics was confined to a few activities. As such, most of the experts gave a narrow
definition of it. However, over a long period of time as its role gradually expanded,
Statistics came to be considered as much wider in its scope and, accordingly, the experts
gave a wider definition of it.
Spiegal, for instance, defines Statistics, highlighting its role in decision-making
particularly under uncertainty, as follows:
2. Objectives of the Study: We should know what the objectives of the proposed
study are. We should ensure that the objectives are not extremely ambitious or else the
study may fail to achieve them because of limitations of time, finance or even
competence of those conducting the study.
3. Determining Sources of Data: The problem and the objectives, thus properly
understood, will enable us to know as to what data are required to conduct the study.
We have to decide whether we should collect primary data or depend exclusively on
secondary data. Sometimes the study is based on both the secondary and the primary
data. When study is to be based on secondary data, whether partly or fully, it is
necessary to ensure that the data are quite suitable and adequate for the objectives of
the study.
4. Designing Data Collection Forms: Once the decision in favour of collection of
primary data is taken, one has to decide the mode of their collection. The two methods
available are: (i) observational method, and (ii) survey method. Suitable questionnaire is
to be designed to collect data from respondents in a field survey.
5. Conducting the Field Survey: Side by side when the data collection forms are
being designed, one has to decide whether a census surveyor a sample survey is to be
conducted. For the latter, a suitable sample design and the sample size are to be chosen.
The field survey is then conducted by interviewing sample respondents. Sometimes, the
survey is done by mailing questionnaires to the respondents instead of contacting them
personally.
6. Organising the Data: The field survey provides raw data from the respondents. It
is now necessary to organise these data in the form of suitable tables and charts so that
we may be aware of their salient features.
7. Analysing the Data: On the basis of the preliminary examination of the data
collected as well as the nature and scope of our problem, we have to analyse data. As
several statistical techniques are available, we should take special care to ensure that the
most appropriate technique is selected for this purpose.
8. Reaching Statistical bindings: The analysis in the preceding step will bring out
some statistical findings of the study. Now we have to interpret these findings in terms
of the concrete problem with which we started our investigation.
9. Presentation of Findings: Finally, we have to present the findings of the study,
properly interpreted, in a suitable form. Here, the choice is between an oral presentation
and a written one. In the case of an oral presentation, one has to be extremely selective
in choosing the material, as in a limited time one has to provide a broad idea of the
study as well as its major findings to be understood by the audience in proper
perspective. In case of a written presentation, a report has to be prepared. It should be
reasonably comprehensive and should have graphs and diagrams to facilitate the reader
in understanding it in all its ramifications.
A number of times business firms are interested to know whether there is an association
between two or more variables such as advertising expenditure and sales. In view of
increasing competitiveness, business and industry spend a large amount on advertising.
It is in their interest to find out whether such advertising expenditure promotes the
sales. Here, by using correlation and regression techniques it can be ascertained that the
advertising expenditure is worthwhile of not.
Mutual Funds
Mutual funds which have come into existence in recent years, provide an avenue to a
person to invest his savings so that he may get a reasonably good return. Different
mutual funds have different objectives as they have varying degrees of risk involved in
the companies they invest in. Here, Statistics provides certain tools or techniques to a
consultant or financial adviser through which he can provide sound advice to a
prospective investor.
Relevance in Banking and Insurance Institutions
Banks and insurance companies frequently use varying statistical techniques in their
respective areas of operation. They have to maintain their accounts and analyze these to
examine their performance over a specified period.
The above discussion is only illustrative and there are numerous other areas where the
use of Statistics is so common that without its use they may have to close down their
operations.
STATISTICAL QUALITY CONTROL METHODS
In the sphere of production, for example, statistics can be useful in various ways to
ensure the production of quality goods. This is achieved by identifying and rejecting
defective or substandard goods. The sale targets can be fixed on the basis of sale
forecasts, which are done by using varying methods of forecasting. Analysis of sales
done against the targets set earlier would indicate the deficiency in achievement, which
may be on account of several causes: (i) targets were too high and unrealistic (ii)
salesmen's performance has been poor (iii) emergence of increase in competition, and
(iv) poor quality of company's product, and so on. These factors can be further
investigated.
LIMITATIONS OF STATISTICS
The preceding discussion highlighting the importance of Statistics in business should
not lead anyone to conclude that Statistics is free from any limitation. As we shall see
here, Statistics has a number of limitations.
There are certain phenomena or concepts where Statistics cannot be used. This is
because these phenomena or concepts are not amenable to measurement. For example,
beauty, intelligence, courage cannot be quantified. Statistics has no place in all such
cases where quantification is not possible.
1. Statistics reveal the average behaviour, the normal or the general trend. An
application of the 'average' concept if applied to an individual or a particular situation
may lead to a wrong conclusion and sometimes may be disastrous. For example, one
may be misguided when told that the average depth of a river from one bank to the other
is four feet, when there may be some points in between where its depth is far more than
four feet. On this understanding, one may enter those points having greater depth,
which may be hazardous.
2. Since Statistics are collected for a particular purpose, such data may not be relevant
or useful in other situations or cases. For example, secondary data (i.e., data originally
collected by someone else) may not be useful for the other person.
3. Statistics is not 100 percent precise as is Mathematics or Accountancy. Those who use
Statistics should be aware of this limitation.
4. In Statistical surveys, sampling is generally used as it is not physically possible to
cover all the units or elements comprising the universe. The results may not be
appropriate as far as the universe is concerned. Moreover, different surveys based on
the same size of sample but different sample units may yield different results.
5. At times, association or relationship between two or more variables is studied in
Statistics, but such a relationship does not indicate 'cause and effect' relationship. It
simply shows the similarity or dissimilarity in the movement of the two variables. In
such cases, it is the user who has to interpret the results carefully, pointing out the type
of relationship obtained.
6. A major limitation of Statistics is that it does not reveal all pertaining to a certain
phenomenon.
7. There is some background information that Statistics does not cover. Similarly, there
are some other aspects related to the problem on hand, which are also not covered. The
user of Statistics has to be well informed and should interpret Statistics keeping in mind
all other aspects having relevance on the given problem.
SUMMARY
This chapter outlined the importance and growth of statistics. Various applications of
statistics in the domain of management have been dealt in detail.
KEY TERMS
Statistics
Statistical quality control methods
Seasonal behaviour
IMPORTANT QUESTIONS
1. Define statistics
2. What do you mean by statistical quality methods?
3. Explain the application of statistics in various business domains.
OBJECTIVE
STRUCTURE
A population is any entire collection of people, animals, plants or things from which we
may collect data. It is the entire group we are interested in, which we wish to describe or
draw conclusions about.
In order to make any generalizations about a population, a sample, that is meant to be
representative of the population, is often studied. For each population there are many
possible samples. A sample statistic gives information about a corresponding population
parameter. For example, the sample mean for a set of data would give information about
the overall population mean.
It is important that the investigator carefully and completely defines the population
before collecting the sample, including a description of the members to be included.
Example
The population for a study of infant health might be all children born in the India, in the
1980's. The sample might be all babies born on 7th May in any of the years.
Population Parameters
Parameter is a value, usually unknown (and which therefore has to be estimated), used
to represent a certain population characteristic. For example, the population mean is a
parameter that is often used to indicate the average value of a quantity.
Within a population, a parameter is a fixed value which does not vary. Each sample
drawn from the population has its own value of any statistic that is used to estimate this
parameter. For example, the mean of the data in a sample is used to give information
about the overall mean in the population from which that sample was drawn.
MEASURES OF CENTRAL LOCATION (OR CENTRAL TENDENCY)
The most important objective of statistical analysis is to determine a single value for the
entire mass of data, which describes the overall level of the group of observations and
can be called a representative set of data. It tells us where the centre of the distribution
of data is located on the scale that we are using. There are several such measures, but we
shall discuss only those that are most commonly used. These are: Arithmetic Mean,
Mode and Median. These values are very useful in not only presenting the overall
picture of the entire data, but also for the purpose of making comparisons among two or
more sets of data.
As an example, questions like, "How hot is the month of June in Mumbai?" can be
answered, generally, by a single figure of the average temperature for that month. For
the purpose of comparison, suppose that we want to find out if boys and girls at the age
10 differ in height. By taking the average height of boys of that age and the average
height of the girls of the same age, we can compare and note the difference.
While, arithmetic mean is the most commonly used measure of central location, mode
and median arc more suitable measures under certain set of conditions and for certain
types of data. However, all measures of central tendency should meet the following
requisites:
really typical of the entire series. Hence, the average chosen should be such that it
is not unduly influenced by extreme values.
Let us consider these three measures of the central tendency:
MODE
In statistics, mode means the most frequent value assumed by a random variable, or
occurring in a sampling of a random variable. The term is applied both to probability
distributions and to collections of experimental data.
Like the statistical mean and the median, die mode is a way of capturing important
information about a random variable or a population in a single quantity. The mode is in
general different from mean and median, and may be very different for strongly skewed
distributions.
The mode is not necessarily unique, since the same maximum frequency may be
attained at different values. The worst case is given by so-called uniform distributions,
in which all values are equally likely.
Mode of a probability distribution
The mode of a probability distribution is the value at which its probability density
function attains its maximum value, so, informally speaking; the mode is at the peak.
Mode of a sample
The mode of a data sample is the element that occurs most often in the collection. For
example, the mode of the sample [1, 3, 6, 6, 6, 6, 7, 7, 12, 12, 17] is 6. Given the list of
data [1, 1, 2, 4, 4] the mode is not unique.
For a sample from a continuous distribution, such as [0.935..., 1.211..., 2.430..., 3.668...,
3.874...], the concept is unusable in its raw form, since each value will occur precisely
once. The usual practice is to discreteise the data by assigning the values to equidistant
intervals, as for making a histogram, effectively replacing the values by the midpoints of
the intervals they are assigned to. The mode is then the value where the histogram
reaches its peak. For small or middle-sized samples the outcome of this procedure is
sensitive to the choice of interval width if chosen too narrow or too wide; typically one
should have a sizable fraction of the data concentrated in a relatively small number of
intervals (5 to 10), while the fraction of the data falling outside these intervals is also
sizable.
Comparison of mean, median and mode
For a probability distribution, the mean is also called the expected value of the random
variable. For a data sample, the mean is also called the average.
In continuous uni-modal distributions the median lies, as a rule of thumb, between the
mean and the mode, about one third of the way going from mean to mode. In a formula,
median = (2 x mean + mode)/3. This rule, due to Karl Pearson, is however not a hard
and fast rule. It applies to distributions that resemble a normal distribution.
Example for a skewed distribution
A well-known example of a skewed distribution is personal wealth: Few people are very
rich, but among those some are excessively rich. However, many are rather poor.
A well-known class of distributions that can be arbitrarily skewed is given by the lognormal distribution. It is obtained by transforming a random variable X having a
normal distribution into random variable Y = exp(X). Then the logarithm of random
variable Y is normally distributed, whence the name.
Taking the mean of X to be 0, the median of Y will be 1, independent of the standard
deviation of X. This is so because X has a symmetric distribution, so its median is also
0. The transformation from X to Y is monotonic, and so we find the median exp(0) = 1
for Y.
When X has standard deviation = 0.2, the distribution of Y is not very skewed. We find
(see under Log-normal distribution), with values rounded to four digits:
Mean = 1.0202
Mode = 0.9608
Indeed, the median is about one third on the way from mean to mode.
When X has a much larger standard deviation, = 5, the distribution of Y is strongly
skewed. Now
Mean = 7.3891
Mode = 0.0183
Here, Pearson's rule of thumb fails miserably.
MEDIAN
In probability theory and statistics, a median is a number dividing the higher half of a
sample, a population, or a probability distribution from the lower half. The median of a
finite list of numbers can be found by arranging all the observations from lowest value to
highest value and picking the middle one. If there is an even number of observations,
one often takes the mean of the two middle values.
At most, half the population has values less than the median and at most half have
values greater than the median. If both groups contain less than half the population,
then some of the population is exactly equal to the median.
Popular explanation
The difference between the median and mean is illustrated in a simple example.
Suppose 19 paupers and one billionaire are in a room. Everyone removes all money from
their pockets and puts it on a table. Each pauper puts $5 on the table; the billionaire
puts $1 billion (that is, $109) there. The total is then $1,000,000,095. If that money is
divided equally among the 20 persons, each gets $50,000,004.75. That amount is the
mean (or "average") amount of money that the 20 persons brought into the room. But
the median amount is $5, since one may divide the group into two groups of 10 persons
each, and say that everyone in the first group brought in no more than $5, and each
person in the second group brought in no less than $5. In a sense, the median is the
amount that the typical person brought in. By contrast, the mean (or "average") is not at
all typical, since no one present - pauper or billionaire - brought in an amount
approximating $50,000,004.75.
Non-uniqueness
There may be more than one median. For example if there are an even number of cases,
and the two middle values are different, then there is no unique middle value. Notice,
however, that at least half the numbers in the list are less than or equal to either of the
two middle values, and at least half are greater than or equal to either of the two values,
and the same is true of any number between the two middle values. Thus either of the
two middle values and all numbers between them are medians in that case.
Measures of statistical dispersion
When the median is used as a location parameter in descriptive statistics, there are
several choices for a measure of variability: the range, the inter-quartile range, and the
absolute deviation. Since the median is the same as the second quartile, its calculation is
illustrated in the article on quartiles. To obtain the median of an even number of
numbers, find the average of the two middle terms.
Medians of particular distributions
The median of a normal distribution with mean and variance 2 is . In fact, for a
normal distribution, mean = median = mode.
The median of a uniform distribution in the interval [a, b] is (a + b) / 2, which is also the
mean.
Medians in descriptive statistics
The median is primarily used for skewed distributions, which it represents differently
than the arithmetic mean. Consider the multiset {1, 2, 2, 2, 3, 9}. The median is 2 in this
case, as is the mode, and it might be seen as a better indication of central tendency than
the arithmetic mean of 3.166....
Calculation of medians is a popular technique in summary statistics and summarizing
statistical data, since it is simple to understand and easy to calculate, while also giving a
measure that is more robust in the presence of outlier values than is the mean.
MEAN
In statistics, mean has two related meanings:
- The average in ordinary English, which is also called the arithmetic mean (and is
distinguished from the geometric mean or harmonic mean). The average is also called
sample mean. The expected value of a random variable, which is also called the
population mean.
- In statistics, means are often used in geometry and analysis. A wide range of means
have been developed for these purposes, which are not much used in statistics. See the
other means section below for a list of means.
Sample mean is often used as an estimator of the central tendency such as the
population mean. However, other estimators are also used.
For a real-valued random variable X, the mean is the expectation of X. If the expectation
does not exist, then the random variable has no mean.
For a data set, the mean is just the sum of all the observations divided by the number of
observations. Once we have chosen this method of describing the communality of a data
set, we usually use the standard deviation to describe how the observations differ. The
standard deviation is the square root of the average of squared deviations from the
mean.
The mean is the unique value about which the sum of squared deviations is a minimum.
If you calculate the sum of squared deviations from any other measure of central
tendency, it will be larger than for the mean. This explains why the standard deviation
and the mean are usually cited together in statistical reports.
An alternative measure of dispersion is the mean deviation, equivalent to the average
absolute deviation from the mean. It is less sensitive to outliers, but less tractable when
combining data sets
Arithmetic Mean
The arithmetic mean is the "standard" average, often simply called the "mean".
The mean may often be confused with the median or mode. The mean is the arithmetic
average of a set of values, or distribution; however, for skewed distributions, the mean is
not necessarily the same as the middle value (median), or most likely (mode). For
example, mean income is skewed upwards by a small number of people with very large
incomes, so that the majority has an income lower than the mean. By contrast, the
median income is the level at which half the population is below and half is above. The
mode income is the most likely income, and favors the larger number of people with
lower incomes. The median or mode is often more intuitive measures of such data.
That said, many skewed distributions are best described by their mean - such as the
Exponential and Poisson distributions.
An amusing example
Most people have an above average number of legs. The mean number of legs is going to
be less than 2, because there are people with one leg, people with no legs and no people
with more than two legs. So since most people have two legs, they have an above average
number.
Geometric Mean
The geometric mean is an average that is useful for sets of numbers that are interpreted
according to their product and not their sum (as is the case with the arithmetic mean).
For example rates of growth.
For example, the geometric mean of 34, 27, 45, 55, 22, 34 (six values) is (34 x 27 x 45 x
55 x 22 x 34)1/6 = (1699493400)1/6 = 34.545
Harmonic Mean
The harmonic mean is an average which is useful for sets of numbers which are defined
in relation to some unit, for example speed (distance per unit of time).
An example
An experiment yields the following data: 34, 27, 45, 55, 22, 34. We need to find the
harmonic mean. No. of items is 6, therefore n = 6. Value of the denominator in the
formula is 0.181719152307. Reciprocal of this value is 5.50299727522. Now, we multiply
this by n to get the harmonic mean as 33.0179836513.
Weighted Arithmetic Mean
The weighted arithmetic mean is used, if one wants to combine average values .rom
samples of the same population with different sample sizes:
The weights i represent the bounds of the partial sample. In other applications they
represent a measure for the reliability of the influence upon the mean by respective
values.
SUMMARY
This chapter has given the meaning of population parameters. The procedures of
measuring the above population parameters are dealt with in detail in the chapter.
KEYTERMS
Population parameters
Mean, Mode and Median
Arithmetic mean
Geometric mean
Harmonic mean
Skewed distribution
IMPORTANT QUESTIONS
1. Explain the methods to measure the Median, Mode and Mean
- End of Chapter -
LESSON 6
HYPOTHESIS TESTING
OBJECTIVES
STRUCTURE
Hypothesis test
Statistical and practical significance
One and two failed tests
Type I and Type II Errors
Null hypothesis: usually the hypothesis that sample observations result purely
from chance effects.
Alternative hypothesis: the hypothesis that sample observations are influenced by
some non-random cause.
For example, suppose we wanted to determine whether a coin was fair and balanced. A
null hypothesis might be that half the flips would result in Heads and half in Tails. The
alternative hypothesis might be that the number of Heads and Tails would be very
different. Suppose we flipped the coin 50 times resulting in 40 Heads and 10 Tails Given
this result, we would be inclined to reject the null hypothesis and accept the alternative
hypothesis.
HYPOTHESIS TESTS
Statisticians follow a formal process to determine whether to accept or reject a null
hypothesis, based on sample data. This process, called hypothesis testing, consists of
four steps.
1. Formulate the hypotheses. This involves stating the null and alternative
hypotheses. The hypotheses are stated in such a way that they are mutually exclusive.
That is, if one is true, the other must be false, and vice versa.
2. Identify the test statistic. This involves specifying the statistic (e.g., a mean score,
proportion) that will be used to assess the validity of the null hypothesis.
3. Formulate a decision rule. A decision rule is a procedure that the researcher uses
to decide whether to accept or reject the null hypothesis.
4. Accept or reject the null hypothesis. Use the decision rule to evaluate the test
statistic. If the statistic is consistent with the null hypothesis, accept the null hypothesis;
otherwise, reject the null hypothesis.
This section provides an introduction to hypothesis testing. Basic analysis involves some
hypothesis testing. Examples of hypotheses generated in marketing research abound
the experimenter probably expects alcohol to have a harmful effect. If the experimental
data show a sufficiently large effect of alcohol, then the null hypothesis that alcohol has
no effect can be rejected.
It should be stressed that researchers very frequently put forward a null hypothesis in
the hope that they can discredit it. For a second example, consider an educational
researcher who designed a new way to teach a particular concept in science, and wanted
to test experimentally whether this new method worked better than the existing method.
The researcher would design an experiment comparing the two methods. Since the null
hypothesis would be that there is no difference between the two methods, the researcher
would be hoping to reject the null hypothesis and conclude that the method he or she
developed is the better of the two.
The symbol H0 is used to indicate the null hypothesis. For the example just given, the
null hypothesis would be designated by the following symbols: H0: 1 - 2 = 0, or by
H0: 1 = 2
The null hypothesis is typically a hypothesis of no difference as in this example where it
is the hypothesis of no difference between population means. That is why the word
"null" in "null hypothesis" is used - it is the hypothesis of no difference.
Despite the "null" in "null hypothesis", there are occasions when the parameter is not
hypothesized to be 0. For instance, it is possible for the null hypothesis to be that the
difference between population means is a particular value. Or, the null hypothesis could
be that the mean SAT score in some population is 600. The null hypothesis would then
be stated as: H0: = 600. Although the null hypotheses discussed so far have all
involved the testing of hypotheses about one or more population means, null hypotheses
can involve any parameter.
An experiment investigating the correlation between job satisfaction and performance
on the job would test the null hypothesis that the population correlation () is 0.
Symbolically, H0: = 0. Some possible null hypotheses are given below:
H0: = 0
H0: = 10
H0: 1 - 2 = 0
H0: = 0.5
H0: 1- 2 = 0
H0: 1 - 2 - 3
H0: 1 2 = 0
This argument is misguided. Intuitively, there are strong reasons for inferring that the
direction of the difference in the population is the same as the difference in the sample.
There is also a more formal argument. A non significant effect might be described as
follows:
Although subjects in the drug group scored higher (M = 23) than did subjects in the
control group (M = 20), the difference between means was not significant, t(18) = 1.4. p
= 0.179. It would not have been correct to say that there was no difference between the
performances of the two groups. There was a difference. It is just that the difference was
not large enough to rule out chance as an explanation of the difference. It would also
have been incorrect to imply that there is no difference in the population. Be sure not to
accept the null hypothesis.
The Precise Meaning of the Probability Value
There is often confusion about the precise meaning of the probability computed in a
significance test. As stated in Step 4 of the steps in hypothesis testing, the null
hypothesis (H0) is assumed to be true. The difference between the statistic computed in
the sample and the parameter specified by H0 is computed and the probability of
obtaining a difference this large or large is calculated. This probability value is the
probability of obtaining data as extreme or more extreme than the current data
(assuming H0 is true). It is not the probability of the null hypothesis itself. Thus, if the
probability value is 0.005, this does not mean that the probability that the null
hypothesis is true is .005. It means that the probability of obtaining data as different or
more different from the null hypothesis as those obtained in the experiment is 0.005.
The inferential step to conclude that the null hypothesis is false goes, as follows:
The data (or data more extreme) are very unlikely given that the null hypothesis is true.
This means that: (1) a very unlikely event occurred or (2) the null hypothesis is false.
The inference usually made is that the null hypothesis is false.
To illustrate that the probability is not the probability of the hypothesis, consider a test
of a person who claims to be able to predict whether a coin will come up heads or tails.
One should take a rather skeptical attitude toward this claim and require strong
evidence to believe in its validity. The null hypothesis is that the person can predict
correctly half the time (H0: = 0.5). In the test, a coin is flipped 20 times and the person
is correct 11 times. If the person has no special ability (H0 is true), then the probability of
being correct 11 or more times out of 20 is 0.41. Would someone who was originally
skeptical now believe that there is only a 0.41 chance that the null hypothesis is true?
They almost certainly would not since they probably originally thought H0 had a very
high probability of being true (perhaps as high as 0.9999). There is no logical reason for
them to decrease their belief in the validity of the null hypothesis since the outcome was
perfectly consistent with the null hypothesis.
The proper interpretation of the test is as follows:
A person made a rather extraordinary claim and should be able to provide strong
evidence in support of the claim if the claim is to be believed. The test provided data
consistent with the null hypothesis that the person has no special ability since a person
with no special ability would be able to predict as well or better more than 40% of the
time. Therefore, there is no compelling reason to believe the extraordinary claim.
However, the test does not prove the person cannot predict better than chance; it simply
fails to provide evidence that he or she can. The probability that the null hypothesis is
true is not determined by the statistical analysis conducted as part of hypothesis testing.
Rather, the probability computed is the probability of obtaining data as different or
more different from the null hypothesis (given that the null hypothesis is true) as the
data actually obtained.
According to one view of hypothesis testing, the significance level should be specified
before any statistical calculations are performed. Then, when the probability (p) is
computed from a significance test, it is compared with the significance level. The null
hypothesis is rejected if p is at or below the significance level; it is not rejected if p is
above the significance level. The degree to which p ends up being above or below the
significance level does not matter. The null hypothesis either is or is not rejected at the
previously stated significance level. Thus, if an experimenter originally stated that he or
she was using the 0.05 significance level and p was subsequently calculated to be 0.042,
then the person would reject the null hypothesis at the 0.05 level. If p had been 0.0001
instead of 0.042 then the null hypothesis would still be rejected at the 0.05 level. The
experimenter would not have any basis to be more confident that the null hypothesis
was false with a p of 0.0001 than with a p of 0.041. Similarly, if the p had been 0.051
then the experimenter would fail to reject the null hypothesis.
He or she would have no more basis to doubt the validity of the null hypothesis than if p
had been 0.482. The conclusion would be that the null hypothesis could not be rejected
at the 0.05 level. In short, this approach is to specify the significance level in advance
and use p only to determine whether or not the null hypothesis can be rejected at the
stated significance level.
Many statisticians and researchers find this approach to hypothesis testing not only too
rigid, but basically illogical. Who in their right mind would not have more confidence
that the null hypothesis is false with a p of 0.0001 than with a p of 0.042? The less likely
the obtained results (or more extreme results) under the null hypothesis, the more
confident one should be that the null hypothesis is false. The null hypothesis should not
be rejected once and for all. The possibility that it was falsely rejected is always present,
and, all else being equal, the lower the p value, the lower this possibility.
Statistical and Practical Significance
It is important not to confuse the confidence with which the null hypothesis can be
rejected with size of the effect. To make this point concrete, consider a researcher
assigned the task of determining whether the video display used by travel agents for
booking airline reservations should be in color or in black and white. Market research
had shown that travel agencies were primarily concerned with the speed with which
reservations can be made. Therefore, the question was whether color displays allow
travel agents to book reservations faster. Market research had also shown that in order
to justify the higher price of color displays, they must be faster by an average of at least
10 seconds per transaction. Fifty subjects were tested with color displays and 50 subjects
were tested with black and white displays.
Subjects were slightly faster at making reservations on a color display (M = 504.7
seconds) than on a black and white display (M = 508.2) seconds, although the difference
is small, it was statistically significant at the .05 significance level. Box plots of the data
are shown below.
the difference was "more significant" due to the larger sample size used. If the
population difference is zero, then a sample difference of 3.4 or larger with a sample size
of 100 is less likely than a sample difference of 3.5 or larger with a sample size of 50.
The 95% confidence interval on the difference between means is:
-5.8 < colour black & white -0.9
and the 99% interval is:
-6.6 < colour black & white -0.1
Therefore, despite the finding of a "more significant" difference between means the
experimenter can be even more certain that the color displays are only slightly better
than the black and white displays. The second experiment shows conclusively that the
difference is less than 10 seconds.
This example was used to illustrate the following points:
(1) an effect that is statistically significant is not necessarily large enough to be of
practical significance and
(2) the smaller of two effects can be "more significant" than the larger.
Be careful how you interpret findings reported in the media. If you read that a particular
diet lowered cholesterol significantly, this does not necessarily mean that the diet
lowered cholesterol enough to be of any health value. It means that the effect on
cholesterol in the population is greater than zero.
TYPE I AND II ERRORS
There are two kinds of errors that can be made in significance testing:
(1) a true null hypothesis can be incorrectly rejected and
(2) a false null hypothesis can fail to be rejected.
The former error is called a Type I error and the latter error is called a Type II error.
These two types of errors are defined in the table.
The probability of a Type I error is designated by the Greek letter alpha () and is called
the Type I error rate; the probability of a Type II error is designated by the Greek
letter beta (), and is called the Type II error rate.
A Type II error is only an error in the sense that an opportunity to reject the null
hypothesis correctly was lost. It is not an error in the sense that an incorrect conclusion
was drawn since no conclusion is drawn when the null hypothesis is not rejected. A Type
I error, on the other hand, is an error in every sense of the word. A conclusion is drawn
that the null hypothesis is false when, in fact, it is true. Therefore, Type I errors are
generally considered more serious than Type II errors. The probability of a Type I error
() is called the significance level and is set by the experimenter. There is a tradeoff
between Type I and Type II errors. The more an experimenter protects himself or
herself against Type I errors by choosing a low level, the greater the chance of a Type II
error. Requiring very strong evidence to reject the null hypothesis makes it very unlikely
that a true null hypothesis will be rejected. However, it increases the chance that a false
null hypothesis will not be rejected, thus lowering power. The Type I error rate is almost
always set at .05 or at .01, the latter being more conservative since it requires stronger
evidence to reject the null hypothesis at the .01 level then at the .05 level.
One and Two Tailed Tests
In the section on "Steps in hypothesis testing", the fourth step involves calculating the
probability that a statistic would differ as much or more from parameter specified in the
null hypothesis as does the statistic obtained in the experiment. This statement implies
that a difference in either direction would be counted. That is, if the null hypothesis
were H0: 1 - 2 = 0, and the value of the statistic M1 - M2 were +5, then the probability
of M1 - M2 differing from zero by five or more (in either direction) would be computed.
In other words, probability value would be the probability that either M1 - M2 5 or M1 M2 -5.Assume that the figure shown below is the sampling distribution of M1 - M2.
The figure shows that the probability of a value of +5 or more is 0.036 and that the
probability of a value of -5 or less is .036. Therefore the probability of a value either
greater than or equal to +5 or less than or equal to -5 is 0.036 + 0.036 = 0.072.
A probability computed considering differences in both directions is called a "twotailed" probability. The name makes sense since both tails of the sampling distribution
are considered. There are situations in which an experimenter is concerned only with
differences in one direction. For example, an experimenter may be concerned with
whether or not 1 - 2 is greater than zero. However, if 1 - 2 is not greater than zero,
the experimenter may not care whether it equals zero or is less than zero. For instance, if
a new drug treatment is developed, the main issue is whether or not it is better than a
placebo. If the treatment is not better than a placebo, then it will not be used. It does not
really matter whether or not it is worse than the placebo.
When only one direction is of concern to an experimenter, then a "one-tailed" test can
be performed. If an experimenter were only to be concerned with whether or not 1 - 2
is greater than zero, then the one-tailed test would involve calculating the probability of
obtaining a statistic as greater than the one obtained in the experiment.
In the example, the one-tailed probability would be the probability of obtaining a value
of M1 - M2 greater than or equal to five given that the difference between population
means is zero.
The shaded area in the figure is greater than five. The figure shows that the one-tailed
probability is 0.036.
It is easier to reject the null hypothesis with a one-tailed than with a two-tailed test as
long as the effect is in the specified direction. Therefore, one-tailed tests have lower
Type II error rates and more power than do two-tailed tests. In this example, the onetailed probability (0.036) is below the conventional significance level of 0.05 whereas
the two-tailed probability (0.072) is not. Probability values for one-tailed tests are one
half the value for two-tailed tests as long as the effect is in the specified direction.
One-tailed and two-tailed tests have the same Type I error rate. One-tailed tests are
sometimes used when the experimenter predicts the direction of the effect in advance.
This use of one-tailed tests is questionable because the experimenter can only reject the
null hypothesis if the effect is in the predicted direction. If the effect is in the other
direction, then the null hypothesis cannot be rejected no matter how strong the effect is.
A skeptic might question whether the experimenter would really fail to reject the null
hypothesis if the effect were strong enough in the wrong direction. Frequently the most
interesting aspect of an effect is that it runs counter to expectations. Therefore, an
experimenter who committed himself or herself to ignoring effects in one direction may
be forced to choose between ignoring a potentially important finding and using the
techniques of statistical inference dishonestly. One-tailed tests are not used frequently.
Unless otherwise indicated, a test should be assumed to be two-tailed.
Confidence Intervals & Hypothesis Testing
There is an extremely close relationship between confidence intervals and hypothesis
testing. When a 95% confidence interval is constructed, all values in the interval are
considered plausible values for the parameter being estimated. Values outside the
interval are rejected as relatively implausible. If the value of the parameter specified by
the null hypothesis is contained in the 95% interval then the null hypothesis cannot be
rejected at the 0.05 level. If the value specified by the null hypothesis is not in the
interval then the null hypothesis can be rejected at the 0.05 level. If a 99% confidence
interval is constructed, then values outside the interval are rejected at the 0.01 level.
Imagine a researcher wishing to test the null hypothesis that the mean time to respond
to an auditory signal is the same as the mean time to respond to a visual signal. The null
hypothesis therefore is: visual auditory = 0.
Ten subjects were tested in the visual condition and their scores (in milliseconds) were:
355, 421, 299, 460, 600, 580, 474, 511, 550, and 586.
Ten subjects were tested in the auditory condition and their scores were: 275, 320, 278,
360, 430, 520, 464, 311, 529, and 326.
The 95% confidence interval on the difference between means is: 9 visual
auditory 196.
Therefore only values in the interval between 9 and 196 are retained as plausible values
for the difference between population means. Since zero, the value specified by the null
hypothesis, is not in the interval, the null hypothesis of no difference between auditory
and visual presentation can be rejected at the 0.05 level. The probability value for this
example is 0.034. Any time the parameter specified by a null hypothesis is not contained
in the 95% confidence interval estimating that parameter, the null hypothesis can be
rejected at the 0.05 level or less. Similarly, if the 99% interval does not contain the
parameter then the null hypothesis can be rejected at the 0.01 level. The null hypothesis
is not rejected if the parameter value specified by the null hypothesis is in the interval
since the null hypothesis would still be plausible.
However, since the null hypothesis would be only one of an infinite number of values in
the confidence interval, accepting the null hypothesis is not justified. There are many
arguments against accepting the null hypothesis when it is not rejected. The null
hypothesis is usually a hypothesis of no difference. Thus null hypothesis such as:
1 - 2 = 0
1 - 2 = 0
in which the hypothesized value is zero are most common. When the hypothesized value
is zero then there is a simple relationship between hypothesis testing and confidence
intervals:
If the interval contains zero then the null hypothesis cannot be rejected at the stated
level of confidence. If the interval does not contain zero then the null hypothesis can be
rejected.
This is just a special case of the general rule stating that the null hypothesis can be
rejected if the interval does not contain the hypothesized value of the parameter and
cannot be rejected if the interval contains the hypothesized value. Since zero is
contained in the interval, the null hypothesis that 1 - 2 = 0 cannot be rejected at the
0.05 level since zero is one of the plausible values of 1 - 2. The interval contains both
positive and negative numbers and therefore 1 may be either larger or smaller than 2.
None of the three possible relationships between 1 and 2:
1 - 2 = 0,
1 - 2 > 0, and
1 - 2 < 0
can be ruled out. The data are very inconclusive. Whenever a significance test fails to
reject the null hypothesis, the direction of the effect (if there is one) is unknown.
Now, consider the 95% confidence interval:
6 < 1 - 2 15
Since zero is not in the interval, the null hypothesis that 1 - 2 = 0 can be rejected at the
0.05 level. Moreover, since all the values in the interval are positive, the direction of the
effect can be inferred: 1 > 2.
Whenever a significance test rejects the null hypothesis that a parameter is zero, the
confidence interval on that parameter will not contain zero. Therefore either all the
values in the interval will be positive or all the values in the interval will be negative. In
either case, the direction of the effect is known.
Define the Decision Rule and the Region of Acceptance
The decision rule consists of two parts: (1) a test statistic and (2) a range of values, called
the region of acceptance. The decision rule determines whether a null hypothesis is
accepted or rejected. If the test statistic falls within the region of acceptance, the null
hypothesis is accepted; otherwise, it is rejected.
We define the region of acceptance in such a way that the chance of making a Type I
error is equal to the significance level. Here is how that is done:
Given the significance level , find the upper limit (UL) of the range of acceptance.
There are three possibilities, depending on the form of the null hypothesis i. If the null hypothesis is < M: The upper limit of the region of acceptance will be
equal to the value for which the cumulative probability of the sampling distribution is
equal to one minus the significance level. That is, P(x < UL) = 1 - .
ii. If the null hypothesis is = M: The upper limit of the region of acceptance will
be equal to the value for which the cumulative probability of the sampling distribution is
equal to one minus the significance level divided by 2. That is, P(x < UL) = 1 - /2.
iii. If the null hypothesis is > M: The upper limit of the region of acceptance is
equal to plus infinity.
In a similar way, we find the lower limit (LL) of the range of acceptance. Again, there
are three possibilities, depending on the form of the null hypothesis.
i. If the null hypothesis is < M: The lower limit of the region of acceptance is
equal to minus infinity.
ii. If the null hypothesis is = M: The lower limit of the region of acceptance will be
equal to the value for which the cumulative probability of the sampling distribution is
equal to the significance level divided by 2. That is, P(x < LL) = /2
iii. If the null hypothesis is > M: The lower limit of the region of acceptance will
be equal to the value for which the cumulative probability of the sampling distribution is
equal to the significance level. That is, P(x < LL) =
The region of acceptance is defined by the range between LL and UL.
Accept or Reject the Null Hypothesis
Once the region of acceptance is defined, the null hypothesis can be tested against
sample data. Simply compute the test statistic. In this case, the test statistic is the
sample mean. If the sample mean falls within the region of acceptance, the null
hypothesis is accepted; if not, it is rejected.
Other Considerations
When one tests a hypothesis in the real world, other issues may come into play. Here are
some suggestions that may be helpful.
You will need to make an assumption about the sampling distribution of the mean
score. If the sample is relatively large (i.e., greater than or equal to 30), you can assume,
based on the central limit theorem, that the sampling distribution will be roughly
normal. On the other hand, if the sample size is small (less than 30) and if the
population random variable is approximately normally distributed (i.e., has a bellshaped curve), you can transform the mean score into a t-score. The t-score will have a
t-distribution.
Assume that the mean of the sampling distribution is equal to the test value M
specified in the null hypothesis.
In some situations, you may need to compute the standard deviation of the sampling
distribution sx. If the standard deviation of the population is known, then sx = x
sqrt[(1/n) - (1/N)], where n is the sample size and N is the population size. On the other
hand, if the standard deviation of the population is unknown, then
sx = s x sqrt of [(1/n) - (1/N)], where s is the sample standard deviation.
Example 1
An inventor has developed a new, energy-efficient lawn mower engine. He claims that
the engine will run continuously for 5 hours (300 minutes) on a single gallon of regular
gasoline. Suppose a random sample of 50 engines is tested. The engines run for an
average of 295 minutes, with a standard deviation of 20 minutes. Test the null
hypothesis that the mean run time is 300 minutes against the alternative hypothesis
that the mean run time is not 300 minutes. Use a 0.05 level of significance.
Solution
There are four steps in conducting a hypothesis test, as described in the previous
sections. We work through those steps below:
1. Formulate hypotheses
The first step is to state the null hypothesis and an alternative hypothesis.
Null hypothesis: = 300 minutes
Alternative hypothesis: < > 300 minutes
Note that these hypotheses constitute a two-tailed test. The null hypothesis will be
rejected if the sample mean is too big or if it is too small.
2. Identify the test statistic
In this example, the test statistic is the mean run time of the 50 engines in the sample =
295 minutes.
3. Define the decision rule
The decision rule consists of two parts: (1) a test statistic and (2) a range of values, called
the region of acceptance. We already know that the test statistic is a sample mean equal
to 295. All that remains is to describe the region of acceptance; that is, to define the
lower limit and the upper limit of the region. Here is how that is done.
a. Specify the sampling distribution. Since the sample size is large (greater than
or equal to 30), we assume that the sampling distribution of the mean is normal,
based on the central limit theorem.
b. Define the mean of the sampling distribution. We assume that the mean of the
sampling distribution is equal to the mean value that appears in the null
hypothesis - 300 minutes.
c. Compute the standard deviation of the sampling distribution. Here the
standard deviation of the sampling distribution sx is:
sx = x sqrt[(1/n) - (1/N)]
sx = 20 x sqrt[1/50] = 2.83
where s is the sample standard deviation, n is the sample size, and N is the
population size. In this example, we assume that the population size N is very
large, so that the value 1/N is about zero.
4. Find the lower limit of the region of acceptance
Given a two-tailed hypothesis, the lower limit (LL) will be equal to the value for which
the cumulative probability of the sampling distribution is equal to the significance level
divided by 2. That is, P(x < LL) = /2 = 0.05/2 = 0.025. To find this lower limit, we use
the Normal Distribution table. From table, cumulative probability = 0.025, mean = 300,
and standard deviation = 2.83. The calculation tells us that the lower limit is 294.45,
given those inputs.
a. Find the upper limit of the region of acceptance. Given a two-tailed hypothesis,
the upper limit (UL) will be equal to the value for which the cumulative
probability of the sampling distribution is equal to one minus the significance
level divided by 2. That is, P(x < UL) = 1 - /2 = 1 - 0.025 = 0.975. To find this
upper limit, we use the Normal Distribution Table. From table, cumulative
probability = 0.975, mean = 300, and standard deviation = 2.83. The calculation
tells us that the upper limit is 305.55, given those inputs.
b. Define the mean of the sampling distribution. We assume that the mean of the
sampling distribution is equal to the mean value that appears in the null
hypothesis - 110.
c. Find the lower limit of the region of acceptance. Given a one-tailed hypothesis,
the lower limit (LL) will be equal to the value for which the cumulative
probability of the sampling distribution is equal to the significance level. That is,
P(x < LL) = = 0.01.
To find this lower limit, we use the T-Distribution Table. From table, cumulative
probability = 0.01, population mean = 110, standard deviation = 2.16, and
degrees of freedom = 20 - 1 = 19. The calculation tells us that the sample mean is
104.51, given those inputs. This is the lower limit of our region of acceptance.
d. Find the upper limit of the region of acceptance. Since we have a one-tailed
hypothesis in which the null hypothesis states that the IQ is greater than 110, any
big number is consistent with the null hypothesis. Therefore, the upper limit is
plus infinity.
Thus, we have determined that the range of acceptance is defined by the values between
104.51 and plus infinity.
4. Accept or reject the null hypothesis.
The sample mean in this example was an IQ score of 108. This value falls within the
region of acceptance. Therefore, we cannot reject the null hypothesis that the average IQ
score of students at Bon Air Elementary is equal to 110.
Power of a Hypothesis Test
When we conduct a hypothesis test, we accept or reject the null hypothesis based on
sample data. Because of the random nature of sample data, our decision can have four
possible outcomes.
We may accept the null hypothesis when it is true. Thus, the decision is correct.
We may reject the null hypothesis when it is true. This kind of incorrect decision
is called a Type I error.
We may reject the null hypothesis when it is false. Thus, the decision is correct.
We may accept the null hypothesis when it is false. This kind of incorrect decision
is called a Type II error.
The probability of committing a Type I error is called the significance level and is
denoted by . The probability of committing a Type II error is called Beta and is denoted
by . The probability of not committing a Type II error is called the power of the test.
How to Compute the Power of a Test
When a researcher designs a study to test a hypothesis, he/she should compute the
power of the test (i.e., the likelihood of avoiding a Type II error). Here is how to do that:
1. Define the region of acceptance.
(The process for defining the region of acceptance is described in the previous three
lessons. See, for example, Hypothesis Tests of Mean Score, Hypothesis Tests of
Proportion (Large Sample), and Hypothesis Tests of Proportion (Small Sample).)
2. Specify the critical value.
The critical value is an alternative to the value specified in the null hypothesis. The
difference between the critical value and the value from the null hypothesis is called the
effect size. That is, the effect size is equal to the critical value minus the value from the
null hypothesis.
3. Compute power. Assume that the true population parameter is equal to the critical
value, rather than the value specified in the null hypothesis. Based on that assumption,
compute the probability that the sample estimate of the population parameter will fall
outside the region of acceptance. That probability is the power of the test.
Example 1: Power of the Hypothesis Test of a Mean Score
Two inventors have developed a new, energy-efficient lawn mower engine. One inventor
says that the engine will run continuously for 5 hours (300 minutes) on a single gallon
of regular gasoline. Suppose a random sample of 50 engines is tested. The engines run
for an average of 295 minutes, with a standard deviation of 20 minutes. The inventor
tests the null hypothesis that the mean run time is 300 minutes against the alternative
hypothesis that the mean run time is not 300 minutes, using a 0.05 level of significance.
The other inventor says that the new engine will run continuously for only 290 minutes
on a gallon of gasoline. Find the power of the test to reject the null hypothesis, if the
second inventor is correct.
Solution
The steps required to compute power are presented below.
1. Define the region of acceptance.
Earlier, we showed that the region of acceptance for this problem consists of the values
between 294.45 and 305.55.
2. Specify the critical value.
The null hypothesis tests the hypothesis that the run time of the engine is 300 minutes.
We are interested in determining the probability that the hypothesis test will reject the
null hypothesis, if the true run time is actually 290 minutes. Therefore, the critical value
is 290. Another way to express the critical value is through effect size. The effect size is
equal to the critical value minus the hypothesized value. Thus, effect size is equal to 290
- 300 = -10
3. Compute power.
The power of the test is the probability of rejecting the null hypothesis, assuming that
the true population mean is equal to the critical value. Since the region of acceptance is
294.45 to 305.55, the null hypothesis will be rejected when the sampled run time is less
than 294.45 or greater than 305.55.
Therefore, we need to compute the probability that the sampled run time will be less
than 294.45 or greater than 305.55. To do this, we make the following assumptions:
Given these assumptions, we first assess the probability that the sample run time will be
less than 294.45. This is easy to do, using the Normal Calculator. We enter the following
values into the calculator: value = 294.45; mean = 290, and standard deviation = 2.83.
Given these inputs, we find that the cumulative probability is 0.94207551. This means
the probability that the sample mean will be less than 294.45 is 0.942.
Next, we assess the probability that the sample mean is greater than 305.55. Again, we
use the Normal Calculator. We enter the following values into the calculator: value =
305.55; mean = 290; and standard deviation = 2.83. Given these inputs, we find that the
probability that the sample mean is less than 305.55 (i.e., the cumulative probability) is
0.99999998. Thus, the probability that the sample mean is greater than 305.55 is 1 0.99999998 or 0.00000002.
The power of the test is the sum of these probabilities: 0.94207551 + 0.00000002 =
0.94207553. This means that if the true average run time of the new engine were 290
minutes, we would correctly reject the hypothesis that the run time was 300 minutes
94.2 percent of the time. Hence, the probability of a Type II error would be very small.
Specifically, it would be 1 minus 0.942 or 0.058.
IMPORTANT STATISTICAL DEFINITIONS OF HYPOTHESIS TESTING
Alpha - The significance level of a test of hypothesis that denotes the probability of
rejecting a null hypothesis when it is actually true; In other words, it is the probability of
committing a Type I error.
Hypothesis testing
Null hypothesis
Alternative Hypothesis
Type I error
Type II error
Probability value
One & two tailed tests
Decision rule
Confidence interval and Statistical significance
IMPORTANT QUESTIONS
1. What do you mean by Hypothesis testing?
2. Define: Null hypothesis and Alternative Hypothesis.
3. Write down the steps to be performed for Hypothesis test
4. What are the differences between Type I and Type II error?
5. How will calculate the Probability value?
6. What do you understand by one and two tailed test?
7. Define Decision rule
8. What is the significance of understanding of confidence intervals?
CHI-SQUARE TEST
OBJECTIVES
STRUCTURE
Bivariate tabular analysis
Chi-square requirements
Computing chi-square
Measure of Association
INTRODUCTION
Chi square is a non-parametric test of statistical significance for bivariate tabular
analysis (also known as crossbreaks). Any appropriately performed test of statistical
significance lets you know the degree of confidence you can have in accepting or
rejecting a hypothesis. Typically, the hypothesis tested with chi square is whether or not
two different samples (of people, texts, whatever) are different enough in some
characteristic or aspect of their behavior that we can generalize from our samples, that
the populations from which our samples are drawn are also different in the behavior or
characteristic.
A non-parametric test, like chi square, is a rough estimate of confidence; it accepts
weaker, less accurate data as input than parametric tests (like t-tests and analysis of
variance, for example) and therefore has less status in the pantheon of statistical tests.
Nonetheless, its limitations are also its- strengths; because chi square is more 'forgiving'
in the data it will accept, it can be used in a wide variety of research contexts.
Chi square is used most frequently to test the statistical significance of results reported
in bivariate tables, and interpreting bivariate tables is integral to interpreting the results
of a chi square test, so we'll take a look at bivariate tabular (crossbreak) analysis.
Bivariate Tabular Analysis
Bivariate tabular (crossbreak) analysis is used when you are trying to summarize the
intersections of independent and dependent variables and understand the relationship
(if any) between those variables. For example, if we wanted to know if there is any
relationship between the biological sex of American undergraduates at a particular
university and their footwear preferences, we might select 50 males and 50 females as
randomly as possible, and ask them, "On average, do you prefer to wear sandals,
sneakers, leather shoes, boots, or something else?" In this example, our independent
variable is biological sex (in experimental research, the independent variable is actively
manipulated by the researcher, for example, whether or not a rat gets a food pellet when
it pulls on a striped bar. In most sociological research, the independent variable is not
actively manipulated in this way, but controlled by sampling for, e.g., males vs. females).
Put another way, the independent variable is the quality or characteristic that you
hypothesize helps to predict or explain some other quality or characteristic (the
dependent variable). We control the independent variable (and as much else as possible
and natural) and elicit and measure the dependent variable to test our hypothesis that
there is some relationship between them. Bivariate tabular analysis is good for asking
the following kinds of questions:
1. Is there a relationship between any two variables in the data?
2. How strong is the relationship in the data?
3. What is the direction and shape of the relationship in the data?
4. Is the relationship due to some Is the relationship due to some intervening variable(s)
in the data?
To see any patterns or systematic relationship between biological sex of undergraduates
at University of X and reported footwear preferences, we could summarize our results in
a table like this:
Table 1: Male and Female Undergraduate Footwear Preferences
Sandals Sneakers
Leather
Boots
shoes
Other
Male
Female
Depending upon how our 50 male and 50 female subjects responded, we could make a
definitive claim about the (reported) footwear preferences of those 100 people.
In constructing bivariate tables, typically values on the independent variable are arrayed
on vertical axis, while values on the dependent variable are arrayed on the horizontal
axis. This allows us to read 'across' from hypothetically 'causal' values on the
independent variable to their 'effects', or values on the dependent variable How you
arrange the values on each axis should be guided "iconically" by your research
question/hypothesis. For example, if values on an independent variable were arranged
from lowest to highest value on the variable and values on the dependent variable were
arranged left to right from lowest to highest, a positive relationship would show up as a
rising left to right line. (But remember, association does not equal causation; an
observed relationship between two variables is not necessarily causal.)
Each intersection/cell-of a value on the independent variable and a value on the
independent variable-reports the result of how many times that combination of values
was chosen/observed in the sample being analyzed. (So you can see that crosstabs are
structurally most suitable for analyzing relationships between nominal and ordinal
variables. Interval and ratio variables will have to first be grouped before they can "fit"
into a bivariate table.) Each cell reports, essentially, how many subjects/observations
produced that combination of independent and dependent variable values? So, for
example, the top left cell of the table above answers the question: "How many male
undergraduates at University of X prefer sandals?" (Answer: 6 out of the 50 sampled.)
Table 1b: Male and Female Undergraduate Footwear Preferences
Sandals Sneakers
Male
17
Female
13
Leather
Boots
shoes
13
9
7
Other
5
16
Reporting and interpreting cross tabs is the most easily done by converting raw
frequencies (in each cell) into percentages of each cell within the values or categories of
the independent variable. For example, in the footwear preferences table above, total
each row, then divide each cell by its own total, and multiply that faction by 100.
Table 1c: Male and Female Undergraduate Footwear Preferences
(percentages)
Sandals Sneakers
Male
12
34
Female
26
10
Leather
Boots
shoes
26
18
14
32
Other
10
50
18
50
column in your data to a hypothetical 100 subjects/observations. If the raw row total
was 93, then percentages do little violence to the raw scores; but if the raw total is 9,
then the generalization (on no statistical basis, i.e., with no knowledge of samplepopulation representativeness) is drastic. So you should provide that total N at the end
of each row/independent variable category (for replicability and to enable the reader to
assess your interpretation of the table's meaning).
With this caveat in mind, you can compare the patterns of distribution of
subjects/observations along the dependent variable between the values of the
independent variable: e.g., compare male and female undergraduate footwear
preference. (For some data, plotting the results on a line graph can also help you
interpret the results: i.e., whether there is a positive (/), negative (), or curvilinear (/, /)
relationship between the variables). Table 1c shows that within our sample, roughly
twice as many females preferred sandals and boots as males, and within our sample,
about three times as many men preferred sneakers as women and twice as many men
preferred leather shoes. We might also infer from the 'Other' category that female
students within our sample had a broader range of footwear preferences than did male
students.
Generalizing from Samples to Populations
Converting raw observed values or frequencies into percentages does allow us to see
more easily patterns in the data, but that is all we can see: what is in the data.
Knowing with great certainty the footwear preferences of a particular group of 100
undergraduates at University of X is of limited use to us; we usually want to measure a
sample in order to know something about the larger populations from which our
samples were drawn. On the basis of raw observed frequencies (or percentages) of a
samples behavior or characteristics, we can make claims about the sample itself, but we
cannot generalize to make claims about the population from which we drew our sample,
unless we submit our results to a test of statistical significance. A test of statistical
significance tells us how confidently we can generalize to a larger (unmeasured)
population from a (measured) sample of that population.
How does chi square do this? Basically, the chi square test of statistical significance is a
series of mathematical formulas which compare the actual observed frequencies of some
phenomenon (in our sample) with the frequencies we would expect if there were no
relationship at all between the two variables in the larger (sampled) population. That is,
chi square tests our actual results against the null hypothesis and assesses whether the
actual results are different enough to overcome a certain probability that they are due to
sampling error. In a sense, chi-square is a lot like percentages; it extrapolates a
population characteristic (a parameter) from the sampling characteristic (a statistic)
similarly to the way percentage standardizes a frequency to a total column N of 100. But
chi-square works within the frequencies provided by the sample and does not inflate (or
minimize) the column and row totals.
Chi Square Requirements
As mentioned before, chi square is a nonparametric test. It does not require the sample
data to be more or less normally distributed (as parametric tests like t-tests do),
although it relies on the assumption that the variable is normally distributed in the
population from which the sample is drawn.
But chi square does have some requirements:
1. The sample must be randomly drawn from the population.
As with any test of statistical significance, your data must be from a random sample of
the population to which you wish to generalize your claims.
2. Data must be reported in raw frequencies (not percentages).
You should only use chi square when your data are in the form of raw frequency counts
of things in two or more mutually exclusive and exhaustive categories. As discussed
above, converting raw frequencies into percentages standardizes cell frequencies as if
there were 100 subjects/observations in each category of the independent variable for
comparability. Part of the chi square mathematical procedure accomplishes this
standardizing, so computing the chi square of percentages would amount to
standardizing an already standardized measurement
3. Measured variables must be independent.
Any observation must fall into only one category or value on each variable. In our
footwear example, our data are counts of male versus female undergraduates expressing
a preference for five different categories of footwear. Each observation/subject is
counted only once, as either male or female (an exhaustive typology of biological sex)
and as preferring sandals, sneakers, leather shoes, boots, or other kinds of footwear. For
some variables, no 'other' category may be needed, but often 'other' ensures that the
variable has been exhaustively categorized. (For some kinds of analysis, you may need to
include an "uncodable" category.) In any case, you must include the results for the whole
sample.
4. Values/categories on independent and dependent variables must be mutually
exclusive and exhaustive.
Furthermore, you should use chi square only when observations are independent: i.e.,
no category or response is dependent upon or influenced by another. (In linguistics,
often this rule is fudged a bit. For example, if we have one dependent variable/column
for linguistic feature X and another column for number of words spoken or written
(where the rows correspond to individual speakers/texts or groups of speakers/texts
which are being compared), there is clearly some relation between the frequency of
feature X in a text and the number of words in a text, but it is a distant, not immediate
dependency.)
5. Observed frequencies cannot be too small.
Lets walk through the process by which a chi square value is computed, using Table 1b
above (renamed Table 1d below).
The first step is to determine our threshold of tolerance for error. That is, what odds are
we willing to accept that we are wrong in generalizing from the results in our sample to
the population it represents? Are we willing to stake a claim on a 50 percent chance that
we're wrong? a 10 percent chance? a five percent chance? 1 percent? The answer
depends largely on our research question and the consequences of being wrong. If
people's lives depend on our interpretation of our results, we might want to take only 1
chance in 100,000 (or 1,000,000) that we're wrong. But if the stakes are smaller, for
example, whether or not two texts use the same frequencies of some linguistic feature
(assuming this is not a forensic issue in a capital murder case!), we might accept a
greater probability--1 in 100 or even 1 in 20 - that our data do not represent the
population we're generalizing about. The important thing is to explicitly motivate your
threshold before you perform any test of statistical significance, to minimize any
temptation for post hoc compromise of scientific standards. For our purposes, we'll set a
probability of error threshold of 1 in 20, or p < .05, for our Footwear study).
The second step is to total all rows and columns.
Table 1d: Male and Female Undergraduate Footwear Preferences: Observed
Frequencies with Row and Column Totals
Sandals Sneakers
Leather
Boots
shoes
13
9
Other
Total
50
Male
17
Female
13
16
50
Total
19
22
20
25
14
100
Remember, that chi square operates by comparing the actual, or observed frequencies in
each cell in the table to the frequencies we would expect if there were no relationship at
all between the two variables in the populations from which the sample is drawn. In
other words, chi square compares what actually happened to what hypothetically would
have happened if all other things were equal' (basically, the null hypothesis). If our
actual results are sufficiently different from the predicted null hypothesis results, we can
reject the null hypothesis and claim that a statistically significant relationship exists
between our variables.
Chi square derives a representation of the null hypothesisthe 'all other things being
equal' scenarioin the following way. The expected frequency in each cell is the product
of that cell's row total multiplied by that cell's column total, divided by the sum total of
all observations. So, to derive the expected frequency of the "Males who prefer Sandals"
cell, we multiply the top row total (50) by the first column total (19) and divide that
product by the sum total 100: (50 X 19)/100 = 9.5. The logic of this is that we are
deriving the expected frequency of each cell from the union of the total frequencies of
the relevant values on each variable (in this case, Male and Sandals), as a proportion of
all observed frequencies (across all values of each variable). This calculation is
performed to derive the expected frequency of each cell, as shown in Table 1e below (the
computation for each cell is listed below:
(Notice that because we originally obtained a balanced male/female sample, our male
and female expected scores are the same. This usually will not be the case.)
We now have a comparison of the observed results versus the results we would expect if
the null hypothesis were true. We can informally analyze this table, comparing observed
and expected frequencies in each cell (Males prefer sandals less than expected), across
values on the independent variable (Males prefer sneakers more than expected, Females
less than expected), or across values on the dependent variable (Females prefer sandals
and boots more than expected, but sneakers and shoes less than expected). But so far,
the extra computation doesn't really add much more information than interpretation of
the results in percentage form. We need some way to measure how different our
observed results are from the null hypothesis. Or, to put it another way, we need some
way to determine whether we can reject the null hypothesis, and if we can, with what
degree of confidence that we're not making a mistake in generalizing from our sample
results to the larger population.
Logically, we need to measure the size of the difference between the pair of observed
and expected frequencies in each cell. More specifically, we calculate the difference
between the observed and expected frequency in each cell, square that difference, and
then divide that product by the difference itself. The formula can be expressed as:
(O - E)2 / E
Squaring the difference ensures a positive number, so that we end up with an absolute
value of differences. If we didn't work with absolute values, the positive and negative
differences across the entire table would always add up to 0. (You really understand the
logic of chi square if you can figure out why this is true.) Dividing the squared difference
by the expected frequency essentially removes the expected frequency from the
equation, so that the remaining measures of observed/expected difference are
comparable across all cells.
So, for example, the difference between observed and expected frequencies for the
Male/Sandals preference is calculated as follows:
1. Observed (6) - Expected (9.5) = Difference (-3.5)
2. Difference (-3.5) squared = 12.25
3. Difference squared (12.25) divided by Expected (9.5) = 1.289
The sum of all products of this calculation on each cell is the total chi square value for
the table. The computation of chi square for each cell is listed below:
(Again, because of our balanced male/female sample, our row totals were the same, so
the male and female observed expected frequency differences were identical. This is
usually not the case)
The total chi square value for Table 1 is 14.026.
Interpreting the Chi Square Value
We now need some criterion or yardstick against which to measure the table's chi square
value, to tell us whether or not it is significant. What we need to know is the probability
of getting a chi square value of a minimum given size even if our variables are not
related at all in the larger population from which our sample was drawn. That is, we
need to know how much larger than 0 (the absolute chi square value of the null
hypothesis) our table's chi square value must be before we can confidently reject the null
hypothesis. The probability we seek depends in part on the degrees of freedom of the
table from which our chi square value is derived.
Degrees of freedom
Mechanically, a table's degrees of freedom (df) can be expressed by the following
formula: df = (r - 1)(c - 1)
That is, a table's degrees of freedom equals the number of rows in the table minus one
multiplied by the number of columns in the table minus one (for 1 x 2 table, df = k - 1,
where k = number of values/categories on the variable).
Degrees of freedom is an issue because of the way in which expected values in each cell
are computed from the row and column totals of each cell. All but one of the expected
values in a given row or column are free to vary (within the total observed-and therefore
expected) frequency of that row or column); once the free to vary expected cells are
specified, the last one is fixed by virtue of the fact that the expected frequencies must
add up to the observed row and column totals (from which they are derived).
Another way to conceive of a table's degrees of freedom is to think of one row and one
column in the table as fixed, with the remaining cells free to vary. Consider the following
visuals (where X = fixed):
df = (r 1) (c 1) = (3 1) (2 1) = 2 x 1 = 2
df = (r 1) (c 1) = (5 1) (3 1) = 4 x 2 = 8
So, for our Table 1,
Table 1: Male and Female Undergraduate Footwear Preferences
Sandals Sneakers
Male
Female
Leather
Boots
shoes
X
X
Other
X
df = (2 - 1) (5 1) = 1 x 4 = 4
In a statistics book, the sampling distribution of chi square (also known as critical values
of chi squares typically listed in an appendix. You read down the column representing
your previously chosen probability of error threshold (e.g., p < 0.5) and across the row
representing the degrees of freedom in your table. If your chi square value is larger than
the critical value in that cell, your data presents a statistically significant relationship
between the variables in your table.
Table 1s chi square value of 14.026 with 4 degrees of freedom, handily clears the related
critical value of 9.49, so we can reject the null hypothesis and affirm the claim that male
and female undergraduates at University of X differ in their (self-reported) footwear
preferences.
Statistical significance does not help you to interpret that nature or explanation of that
relationship; that must be done by other means (including bivariate tabular analysis of
the data). But a statistically significant chi square value does denote the degree of
confidence you may hold that the relationship between variables described in your
results is systematic in the larger population and not attributable to random error.
Statistical significance also does not ensure substantive significance. A large enough
sample may demonstrate a statistically significant relationship between two variables,
but that relationship may be trivially weak one. Statistical significance means only that
the pattern of distribution and relationship between the variables which is found in the
data from a sample can be confidently generalized to the larger population from which
the sample was randomly drawn. By itself it does not ensure that the relationship is
theoretically or practically important, or even very large.
Measures of Association
While the issue of theoretical or practical importance of a statistically significant result
cannot be quantified, the relative magnitude of a statistically significant relationship can
be measured. Chi Square allows you to make decisions about whether there is a
relationship between two or more variables; if the null hypothesis is rejected, we
conclude that there is a statistically significant relationship between the variables. But
we frequently want a measure of the strength of that relationship - an index of degree of
correlation, a measure of the degree of association between the variables represented in
our table (and data). Luckily, several measures of association can be derived from a
tables chi square value.
For tables larger than 2 x 2 (like our Table 1), a measure called 'Cramer's phi' is derived
by the following formula (where N is the total number of observations, and k is the
smaller of the number of rows or columns).
Cramers phi = sqrt [(chi square) / (N x (k-1)]
So, for our Table 1 (2 x 5), Cramers phi will be computed as follows:
1. N (k - 1) = 100 (2 1) = 100
2. Chi square / 100 = 14.026 / 100 = 0.14
3. Square root of 0.14 = 0.37
The product is interpreted as a Pearson r (that is, as a correlation coefficient).
(For 2X2 tables, a measure called 'phi' is derived by dividing the table's chi square value
by N (the total number of observations) and then taking the square root of the product.
Phi is also interpreted as a Pearson r.)
IMPORTANT QUESTIONS
1. What do you mean by bivariate tabular analysis?
2. What are the statistical applications in chi square tests?
3. How will you calculate the degree of freedom?
OBJECTIVES
STRUCTURE
F-test
Cumulative probability and F distribution
F-Test for equality
T-test
Two sample t-test for equal means.
F-TEST
The f statistic, also known as an f value, is a random variable that has an F distribution.
The F Distribution
The distribution of all possible values of the f statistic is called an F distribution, with v1
= n1 1 and v2 = n2 1 degrees of freedom.
The curve of the F distribution depends on the degrees of freedom, v1 and v2. When
describing an F distribution, the number of degrees of freedom associated with the
standard deviation in the numerator of the f statistic is always stated first. Thus, f(5,9)
would refer to an F distribution with v1 = 5 and v2 = 9 degrees of freedom; whereas f(9,5)
would refer to an F distribution with v1 = 9 and v2 = 5 degrees of freedom. Note that the
curve represented by f(5,9) would differ from the curve represented by f(9,5).
The F distribution has the following properties:
The mean of the distribution is equal to v1 / (v2 2)
The variance is equal to
v22 (v1 + 2)
---------------------------v1 (v2 2) (v2 4)
Cumulative Probability and the F Distribution
Every f statistic can be associated with a unique cumulative probability. This cumulative
probability represents the likelihood that the f statistic is less than or equal to a specified
value. Here are the steps required to compute an f statistic:
Suppose you randomly select 7 women from a population of women and 12 men from a
population of men. The table below shows the standard deviation in each sample and in
each population.
Parametric tests provide inferences for making statements about the means of parent
populations. A t-test is commonly used for this purpose. This test is based on the
Student's t statistic. The t statistic assumes that the variable is normally distributed, the
mean is known (or assumed to be known), and the population variance is estimated
from the sample. Assume that the random variable X is normally distributed, with mean
and unknown population variance 2, which is estimated by the sample variance S2.
Recall that the standard deviation of the sample mean, X, is estimated as sx = s/n. Then,
t = (X - ) / sx is distributed with n - 1 degrees of freedom.
The t-distribution is similar to the normal distribution in appearance. Both distributions
are bell-shaped and symmetric. However, as compared to the normal distribution, that
distribution has more area in the tails and less in the center. This is because population
variance 2 is unknown and is estimated by the sample variance S2. Given the
uncertainty in the value of S2, the observed values of t are more variable than those of z.
Thus we must go a large number of standard deviations from 0 to encompass a certain
percentage of values from the t distribution than is the case with the normal
distribution. Yet, as the number of freedom increases, the t distribution approaches the
normal distribution. In large samples of 120 or more, the t distribution and the normal
distribution are virtually indistinguishable. Table 4 in the statistical appendix shows
The two-sample t-test (Snedecor and Cochran, 1989) is used to determine if two
population means are equal. A common application of this is to test if a new process or
treatment is superior to a current process or treatment.
There are several variations on this test:
1. The data may either be paired or not paired. By paired, we mean that there is a oneto-one correspondence between the values in the two samples. That is, if X 1, X2, , Xn
and Y1, Y2... , Yn are the two samples, then Xi corresponds to Yi. For paired samples, the
difference Xi - Yi is usually calculated. For unpaired samples, the sample sizes for the
two samples may or may not be equal. The formulas for paired data are somewhat
simpler than the formulas for unpaired data.
2. The variances of the two samples may be assumed to be equal or unequal. Equal
variances yield somewhat simpler formulas, although with computers this is no longer a
significant issue.
In some applications, you may want to adopt a new process or treatment only if it
exceeds the current treatment by some threshold. In this case, we can state the null
hypothesis in the form that the difference between the two populations means is equal
to some constant (1 2 = d0) where the constant is the desired threshold.
Interpretation of Output
1. We are testing the hypothesis that the population mean is equal for the two samples.
The output is divided into five sections.
2. The first section prints the sample statistics for sample one used in the computation
of the f-test
3. The second section prints the sample statistics for sample two used in the
computation of the t-test.
4. The third section prints the pooled standard deviation, the difference in the means,
the t-test statistic value, the degrees of freedom, and the cumulative distribution
function (cdf) value of the t-test statistic under the assumption that the standard
deviations are equal. The t-test statistic cdf value is an alternative way of expressing the
critical value. This cdf value is compared to the acceptance intervals printed in section
five. For an upper one-tailed test, the acceptance interval is (0,1-), the acceptance
interval for a two-tailed test is (/2, 1- /2), and the acceptance interval for a lower onetailed test is (,1).
5. The fourth section prints the pooled standard deviation, the difference in the means,
the t-test statistic value, the degrees of freedom, and the cumulative distribution
function (cdf) value of the t-test statistic under the assumption that the standard
deviations are not equal. The t-test statistic cdf value is an alternative way of expressing
the critical value. cdf value is compared to the acceptance intervals printed in section
five. For an upper one-tailed test, the alternative hypothesis acceptance interval is (1-,
1), the alternative hypothesis acceptance interval for a lower one-tailed test is (0,), and
the alternative hypothesis acceptance interval for a two-tailed test is (1-/2,1) or (0,/2).
Note that accepting the alternative hypothesis is equivalent to rejecting the null
hypothesis.
6. The fifth section prints the conclusions for a 95% test under the assumption that the
standard deviations are not equal since a 95% test is the most common case." Results
are given in terms of the alternative hypothesis for the two-tailed test and for the onetailed test in both directions. The alternative hypothesis acceptance interval column is
stated in terms of the cdf value printed in section four. The last column specifies
whether the alternative hypothesis is accepted or rejected. For a different significance
level, the appropriate conclusion can be drawn from the t-test statistic cdf value printed
in section four. For example, for a significance level of 0.10, the corresponding
alternative hypothesis acceptance intervals are (0,0.05) and (0.95,1), (0,0.10), and
(0.90,1).
Two-sample f-tests can be used to answer the following questions
1. Is process 1 equivalent to process 2?
2. Is the new process better than the current process?
3. Is the new process better than the current process by at least some pre-determined
threshold amount?
Matrices of t-tests
T-tests for dependent samples can be calculated for long lists of variables, and reviewed
in the form of matrices produced with case wise or pair wise deletion of missing data,
much like the correlation matrices. Thus, the precautions discussed in the context of
correlations also apply to t-test matrices:
a. The issue of artifacts caused by the pair wise deletion of missing data in t-tests and
b. The issue of "randomly" significant test values.
SUMMARY
The above chapter has given the frame work for performing key statistical tests like FTest and T-test. T-test and F-test are parametric tests. T-test is any statistical hypothesis
test in which the test statistic has a Students distribution if the null hypothesis is true.
KEY TERMS
Degrees of Freedom
T test
T distribution
F distribution
Sampling distribution
IMPORTANT QUESTIONS
1. How will you calculate the Degrees of Freedom?
2. Explain the procedures for performing t test?
3. What are the metrics of t test?
4. Explain the procedures for performing f test?
REFERENCE
1. Sumathi, S. and P. Saravanavel - Marketing research and Consumer Behaviour
2. Ferber, R., and Verdoorn, P.J., Research Methods in Business, New York-the
Macmillan Company 1962.
3. Ferber, R., Robert (ed.) Hand Book of Marketing Research. New York McGraw Hill,
Inc. 1948.
OBJECTIVES
To know the different types of Data and the sources of the same.
To learn the different data collection methods and its merits, demerits.
To understand the difference between Questionnaire and Interview Schedule.
To apply the suitable data collection method for the research.
STRUCTURE
Primary data
Secondary data
Interview
Questionnaire
Observation
INTRODUCTION
After defining the research problem and drawing the research design, the important
task of the researcher is Data collection. While deciding on the research method, the
method of data collection to be used for the study also should be planned.
The same of information and the manner in which data are collected could well make a
big difference to the effectiveness of the research project.
In data collection, the researcher should be very clear on what type of data is to be used
for the research. There are two types of data namely primary data and secondary data.
The method of collecting primary and secondary data differ since primary data are to be
originally collected while in case of secondary data, it is merely compilation of the
available data.
PRIMARY DATA
Primary data are generally, information gathered by the researcher for the purpose of
the project at hand. When the data are collected for the first time using experiments and
surveys, the data is known as primary data. So, in case of primary data, it is always the
responsibility of the researcher to decide on further processing of data.
There are several methods of data collection each with its advantages and
disadvantages.
The data collection methods include the following:
1. Interview- Face to face interview, Telephone interview, Computer assigned
interview, Interviews through electronic media
2. Questionnaire - These are personally administered sent through the mail, or
electronically administered
3. Observation - of individuals and events with or without video taping or audio
recording. Hence interviews, questionnaires and observation methods are three main
data collection methods.
Some of the other data collection methods used to collect primary data:
1. Warranty cards
2. Distributor Audits
3. Pantry Audits
4. Consumer Panels
SECONDARY DATA
As already mentioned, it the researcher who decides to collect secondary data for his
research that can be collected through various sources. In the case of secondary data,
the researcher may not face severe problems that are usually associated with primary
data collection.
Secondary data may either be published or unpublished data. Published data may be
available with the following sources:
1. Various publications of the central, state or local governments.
2. Various publications of foreign governments or of international bodies.
3. Technical and Trade Journals.
4. Books, Magazines, Newspapers.
5. Reports and publication from various associations connected with industry and
business.
6. Public records and statistics.
7. Historical documents.
Though there are various sources for secondary data, it is the responsibility of the
researcher that he should make a minute scrutiny of data in order to involve the data
more suitable and adequate.
INTERVIEWS
An interview is a purposeful discussion between two or more people. Interview can help
to gather valid and reliable data. There are several types of interviews.
Types of Interviews
Interviews may be conducted in a very formal manner, using structured and
standardized questionnaire for each respondent.
Interviews also may be conducted informally through unstructured conversation. Based
on formal nature and structure, the interviews are classified as follows:
1. Structured Interviews : These interviews involve the use of a set of predetermined
questions and of highly standardized techniques of recording. So, in this type of
interview a rigid proved method is followed.
Despite the variations in interviews techniques the following are the advantages of the
same.
1. Interviews may help to collect more information and also in depth information.
2. The unstructured interviews are more advantages since, there is always an
opportunity for the interviewers to restructure the questions.
3. Since the respondents are contacted for the information, there are always greater
advantages of creating support and collecting personal information also.
4. Interviews help the researcher to collect all the necessary information, here the
evidence no response will be very low.
5. It is also possible for the interviewer to collect additional information about the
environment, name, behavior and attitude of the respondents.
Disadvantages of interviews
1. Interviews are expensive methods especially in case of widely spread geographical
samples are taken.
2. There may be a possibility for the barriers in the case of both interviewer and the
respondent.
3. These methods are also time-consuming especially when the large samples are taken
for the study.
4. There may be a possibility for the respondent to hide the real opinion, so genuine data
may not be collected.
5. Sometimes there will be great difficult in adopting interview methods became fixing
appointment with the respondent itself may not be possible.
Hence, for successful implementation of the interview method, the interviewer should
be carefully selected, trained and briefed.
How to make interviews successful?
1. As mentioned above, the interviewers must be carefully selected and trained properly.
2. The interviewer should have the knowledge of exploring to collect the needed
information from the respondent.
3. Honesty and integrity of the interviewer also determines the outcomes of the
interview.
The questions in this questionnaire are not structured in advance. These sorts of
questionnaire give more scopes for variety of answers. Mainly to conduct interviews
where in different responses are expected, these type of questionnaire are used.
The researcher should be much clear on when to use the questionnaire. These can be
mostly used in case of descriptive and explanatory type of research. Example:
Questionnaire on attitude, opinion, and organizational practices, enable the researcher
to identify and describe the variability.
Advantages of the Questionnaire
But the rate of return of questionnaire and the fulfillment at needed data for the study
may be doubtful. This can be used only when the respondents are educated and will
know to read the language in which questionnaire is prepared. Possibilities for
ambiguous replies, omission of replies are more. This method is more time consuming.
Before sending the final questionnaire to the respondents it is always more important to
conduct the Pilot study for resting the questionnaire. Pilot study is just a rehearsal of the
main survey such survey conducted with help of experts brings more strength to the
questionnaire.
Data collection through schedules
This method is very much like data collection through questionnaires. Schedules are a
proforma containing set of questions which are filled in by the enumerators who are
specially appointed for this purpose. In this method the enumerators are expected to
perform well and they must be knowledgeable and must possess the capacity of cross
examination in order to find the truth. These methods are usually expensive and are
conducted by bigger and Government organizations.
OBSERVATION
Observation is one of the cheaper and effective techniques of data collection. The
observation is understood as a systematic process of recording the behavioral patterns
of people, objects and occurrence as they are witnessed.
Using this observational method of data collection the following data related to
movements, work habits, the statements mad and meetings conducted,[by human
begins] facial expressions, body language and other emotions such as joy, anger and
sorrowfulness of the human beings can be collected.
The data collected in this method are generally reliable and they are free from
respondent's bias.
It is easier to observe the busy people rather meeting and collecting the data.
The observer has to be present in the situation where the data to be collected.
This method is very slow and expensive.
Since the observer is going to collect the data by consuming more time and
observing the sample there is a possibility for biased information.
Survey Method
Survey is a popular method of "fact-finding" study which involves collection of data
directly from the population at a particular time. Survey can be defined as a research
technique in which information may be gathered from a sample of people by use of
questionnaire or interview.
Survey is considered as field study which is conducted in a natural setting which collects
information directly from the respondents. Surveys are been conducted for many
purposes - population census, socio economic surveys, expenditure surveys and
marketing surveys. The purpose of these surveys would be providing information to the
government planners or business enterprises. The surveys are also conducted to explain
phenomena where in causal relationship between two variables to be assessed.
Surveys are been used to compare the demographic groups such as to compare the high
and low income groups, to compare the preference based on age. These surveys are been
conducted in the care of descriptive type of research and in which large samples are
focused.
The surveys are more appropriate in the social and behavioral science. Surveys are more
convened with formulating the hypothesis and testing the relationship between nonmanipulated variables. The survey research requires skillful workers to gather data.
The subjects for surveys may be classified as:
1. Social surveys which include
2. Economic surveys
Deciding on tools used for data collection
KEY TERMS
Primary data
Secondary data
Interview
Questionnaire
Schedule
Observation
Survey
Warranty cards
Pantry Audits
Distributor Audits
Consumer Panels
QUESTIONS
1. Describe the different data sources, explaining their advantages and disadvantages.
2. What is bias and how it can be reduced during interviews?
3. "Every data collection method has it own built in biases. Therefore resorting to multi
methods of data collection is only going to compound the biases". How would you
evaluate the statement?
4. Discuss the role technology in data collection.
5. What is your view on using the warranty cards and Distributor audits in data
collection?
6. Differentiate the questionnaire and Interview schedules to decide the best one.
7. Discuss the main purposes for which Survey methods are used.
OBJECTIVES
STRUCTURE
Universe/population
Sampling
Sampling techniques
Importance of sample design and sample size
INTRODUCTION
Sampling Methods is a process of selecting sufficient number of elements from the
population. This is also understood as the process of obtaining information about entire
population by examining only a part of it. So sample can be defined as a subset of the
population. In other words, some, but not all elements of the population would form the
sample.
Basically sampling helps a researcher in variety of ways as follows:
It saves time and money. A sample study usually is less expensive than a
population survey.
It also helps the researcher to obtain the accurate results.
Sampling is only way when the population is very large in size.
It enables to estimate the sampling error so; this assists in obtaining information
and in convening the characteristics of population.
To understand the sampling process, the researcher also should understand the
following terms:
1. Universe / Population
In any of the research, the interest of the researcher is mainly in studying the various
characteristics relating to items or individuals belonging to a particular group. This
group of individuals under study is known as the population or universe.
2. Sampling
A finite subset-selected from a population with the objective of investigating its
properties is called a "sample".
The number of units in the sample is known as the 'sample size'. This is the important
role of any research which enables to draw conclusions about characteristics of the
population.
3. Parameter & Statistics
The statistical constants used for further analysis of Data collected such as Mean (),
Variance (2), Skewness (1), Kurtosis (2), Correlation () can be computed for the
sample drawn from the population.
Sampling is the important part of any research before data collection. So the sampling
process should be done in a careful way to obtain the exact samples and sample size of
the population on which the research is to be done.
Example: A researcher who would like to study the customer satisfaction for a health
drink namely Horlicks should identify the population who are consuming Horlicks. If
the consumers are varying in age, genders all over the state or country, he should be able
to decide to particular consumers are going to be focused. Again, if the number is more
to survey he has to decide on how many individuals he targets for his study.
Hence the effecting sampling process should have the following steps:
Define the population
(Elements, units, extent, and time)
Sampling plan
(Procedure for selected sampling unit)
Spatial sampling
Saturation sampling
1. Probability Sampling
This is a scientific technique of drawing samples form the population according to some
laws of change according to which each unit in the universe has some definite preassigned probability of being selected in the sample.
Simple Random Sampling
In this technique, sample is drawn in such a way that every elements or unit in the
population has an equal and independent chance of being included in the sample.
The unit selected in any draw from the population is not preplanned in population
before making the next draw is known as simple random sampling without replacement.
If the unit is replaced back before making the next draw the sampling plan is called as
simple random sampling with replacement.
Stratified Random Sampling
When the population is heterogeneous with respect to the variable or characteristics
under the study this sampling method is used. Stratification means division into
homogenous layers or groups. Stratified random sampling involves stratifying the given
population into a number of sub-groups or sub-population known as strata.
The characteristics of stratified samples are as follows:
Hence using the proportional allocation, the sample size for different strata are 15, 9,
and 6, which are proportionate to the strata sizes of 3000, 1800, and 1200.
Systematic Sampling
This sampling is a slight variation of simple random sampling in which only the first
sample unit is selected at random while remaining units are selected automatically in a
definite sequence at equal spacing from one another. This kind of sampling is
recommended only when if a complete and up to date list of sampling units is available
and the units are arranged in a systematic order as alphabetical, chronological,
geographical etc.
Systematic sampling can be taken as an improvement over a simple random sampling
since it spreads more evenly over the entire population. This method is one of the easier
and less costly methods of sampling and can be conveniently used in case of large
population.
Cluster Sampling
If the total area of interest happens to be a big one, a convenient way in which a sample
can be taken is to divide the area into a number of smaller non-overlapping areas and
then to randomly select a number of these smaller areas.
In cluster sampling, the total population is divided into a number of relatively small sub
divisions which are themselves clusters of still smaller units and some of these clusters
are randomly selected for inclusion in overall sample. Cluster sampling reduces the cost
by concentrating surveys in clusters. But this type of sampling is less accurate than
random sampling.
Multistage Sampling
It is a further development of the principle of cluster sampling. In case of investigating
the working efficiency of nationalized banks in India, a sample of few banks may be
taken for the purpose of investigation.
Here to select the banks as a first step, the main states in a country are selected from the
states from which the banks to be included for the study will be selected. This represents
the two stage sampling. Even further, from the district certain towns may be selected,
from where the banks will be selected. This may represent three stages sampling.
Even thereafter, instead of taking census from all the banks in all the towns we have
selected, once again banks may be selected randomly for the survey. Hence the random
selection at all levels (various levels) is known as multistage random sampling design.
Sequential Sampling
This is one of the complex sampling designs. The ultimate size of sample in this
technique is not fixed in advance but it is determined according to the mathematical
decision rules on the basis of information yielded in the survey.
This method is adopted when sampling plan is accepted in context of Statistical Quality
Control.
Example: When a lot is to be accepted or rejected on the basis of single sample, it is
known as single sampling; when the decision is to be taken on the basis of two
samples it is known as double sampling, and in case the decision is based on the
more than two samples but the number of samples is certain and decided in advance,
the sampling is known as multi sampling. In case when the number of samples is
more than two but it is neither certain nor decided in advance, this type of system is
often referred to as Sequential Sampling. So in case of Sequential Sampling, one can
go on taking samples one after another as long as one desires to do so.
2. Non-probability Sampling
Quota Sampling
This is stratified-cum-purposive or judgment sampling and thus enjoys the benefits of
both. It aims at making the best use of stratification without incurring the high costs
involved in probabilistic methods. There is considerable saving in time and money as
the simple units may be selected that they are close together. If carefully experienced by
skilled and experienced investigators who are aware of the limitations of judgment
sampling and if proper controls are imposed on the investigators, this sampling method
may give reliable results.
Determination of appropriate sample size is crucial part of any business research. The
decision on proper sample size tremendously requires the use of statistical theory. When
a business research report is been evaluated, the evaluation start with the question of
'how big is the sample size?'
Having discussed various sampling designs it is important to focus the attention on
Sample Size. Suppose we select a sample size of 30 from the population of 3000 through
a simple random sampling procedure, will we able to generalize the findings to the
population with confidence? So in this case what is the sample size that would be
required to carry out the research?
It is the known fact that larger the sample size, the more accurate the research is. In fact
this is the fact based on the statistics. According to this fact, increasing the sample size
decreases the width of the confidence interval at a given confidence level. When the
standard deviation of the population is unknown, a confidence interval is calculated by
using the formula:
Confidence Interval,
= X KSx
where,
Sx = S / Sqrt(n)
In sum, choosing the appropriate sampling plan is one of the important research design
decisions the researcher has to make. The choice of a specific design will depend broadly
on the goal of research, the characteristics of the population, and considerations of cost.
Issues of Precision and Confidence in determining Sample size
We now need to focus attention on the second aspect of the sampling design issuethe
sample size. Suppose we select 30 people from a population of 3,000 through a simple
random sampling procedure. Will we be able to generalize our findings to the
population with confidence? What is the sample size that would be required to make
reasonably precise generalizations with confidence? What do precision and confidence
mean?
A reliable and valid sample should enable us to generalize the findings from the sample
to the population under investigation. No sample statistic (X, for instance) is going to be
exactly the same as the population parameter (Sx), no matter how sophisticated the
probability sampling design is. Remember that the very reason for a probability design
is to increase the probability that the sample statistics will be as close as possible to the
population parameters.
Precision
Precision refers to how close our estimate is to the true population characteristic.
Usually, we would estimate the population parameter to fall within a range, based on the
sample estimate.
Example: From a study of a simple random sample of 50 of the total 300 employees in
a workshop, we find that the average daily production rate per person is 50 pieces of a
particular product (X = 50). We might then (by doing certain calculations, as we shall
see later, be able to say that the true average daily production of the product (X) would
lie anywhere between 40 and 60 for the population of employees in the workshop. In
saying this, we offer an interval estimate, within which we expect the true population
mean production to be ( = 50 10). The narrower this interval, the greater is the
precision. For instance, if we are able to estimate that the population mean would fall
anywhere between 45 and 55 pieces of production ( = 50 5) rather than 40 and 60 (
= 50 10), then we would have more precision. That is, we would now estimate the
mean to lie within a narrower range, which in turn means that we estimate with greater
exactitude or precision.
Precision is a function of the range of variability in the sampling distribution of the
sample mean. That is, if we take a number of different samples from a population, and
take the mean of each of these, we will usually find that they are all different, are
normally distributed, and have a dispersion associated with them Even if we take only
one sample of 30 subjects from the population, we will still be able to estimate the
variability of the sampling distribution of the sample mean. This variability is called the
standard error, denoted by 'S'. The standard error is calculated by the following formula:
Sx = S / Srqt(n)
where,
S = Standard deviation of the sample
n = Sample size
Sx = Standard error or the extent of precision offered by the sample.
In sum, the closer we want our sample results to reflect the population characteristics,
the greater will be the precision we would aim at. The greater the precision required, the
larger is the sample size needed, especially when the variability in the population itself is
large.
Confidence
Whereas precision denotes how close we estimate the population parameter based on
the sample statistic, confidence denotes how certain we are that our estimates will really
hold true for the population. In the previous example of production rate, we know we
are more .precise when we estimate the true mean production () to fall somewhere
between 45 and 55 pieces, than somewhere between 40 and 60.
In essence confidence reflects the level of certainty with which we can state that our
estimates of the population parameters, based on our sample statistics will hold true.
The level of confidence can range from 0 to 100%. A 95% confidence is the
conventionally accepted level for most business research, most commonly expressed by
denoting the significance level as p = .05. In other words, we say that at least 95 times
out of 100, our estimate will reflect the true population characteristic.
In sum, the sample size n, is a function of
1. The variability in the population
2. Precision or accuracy needed
3. Confidence level desired
4. Type of sampling plan used, for example, sample random sampling versus stratified
random sampling
It thus becomes necessary for researchers to consider at least four points while making
decisions on the sample size needed to do the research:
(1) Much precision is really needed in estimating the population characteristics interest,
that is, what is the margin of allowable error?
(2) How much confidence is really needed, i.e. how much chance can we take of making
errors in estimating the population parameters?
(3) To what extent is there variability in the population on the characteristics
investigated?
(4) What is the cost-benefit analysis of increasing the sample size?
Determining the Sample Size
Now that we are aware of the fact that the sample size is governed by the extent of
precision and confidence desired, how do we determine the sample, retired for our
research? The procedure can be illustrated through an example:
Suppose a manager wants to be 95% confident that the withdrawals in a bank will be
within a confidence interval of $500. Example of a simple of clients indicates that the
average withdrawals made by them have a standard deviation of $3,500. What would be
the sample size needed in this case?
We noted earlier that the population mean can be estimated by using the formula:
= X KSx
Given, S = 3500. Since the confidence level needed here is 95%, the applicable K value is
1.96 (t-table). The interval estimate of $500 will have to encompass a dispersion of
(1.96 x standard error). That is,
The sample size needed in the above is 188. Let us say that this bank has the total
clientele of only 185. This means we cannot sample 188 clients. We can, in this case,
apply the correction formula and see what sample size would be needed to have the
same level of precision and confidence given the fact that we have a total of only 185
clients. The correction formula is as follows:
where,
N = total number of elements in the population = 185
n = sample size to be estimated = ?
Sx = Standard error of estimate of the mean = 255.10
S = Standard deviation of the sample mean = 3500
Applying the correlation formula,
we find that
255.10 = 3500 n 185-n/184
the value of n to be 94.
We would now sample 94 of the total 185 clients.
To understand the impact of precision and/or confidence on the sample size, let us try
changing the confidence level required in the bank withdrawal exercise which needed a
sample size of 188 for a confidence level of 95%. Let us say that the bank manager now
wants to be 99% sure that the expected monthly withdrawals will be within the interval
of $500. What will be the sample size now needed?
The sample has now to be increased 1.73 times (from 188 to 325) to increase the
confidence level from 95% to 99%.It is hence a good idea to think through how much
precision and confidence one really needs, before determining the sample size for the
research project.
So far we have discussed sample size in the context of precision and confidence with
respect to one variable only. However, in research, the theoretical framework has
several variables of interest, and the question arises how one should come up with a
sample size when all the factors are taken into account.
Krejcie and Morgan (1970) greatly simplified size decision by providing a table that
ensures a good decision model. The Table provides that generalized scientific guideline
for sample size decisions. The interested student is advised to read Krejcie and Morgan
(1970) as well as Cohen (1969) for decisions on sample size.
much as we would be prone to committing Type II errors. Hence, neither too large nor
too small sample sizes help research projects.
Roscoe (1975) proposes the following rules of thumb for determining
sample size:
1. Sample sizes larger than 30 and less than 500 are appropriate for most research.
2. Where samples are to be broken into sub samples; (male/females, juniors/ seniors,
etc.), a minimum sample size of 30 for each category is necessary.
3. In multivariate research (including multiple regression analyses), the sample size
should be several times (preferably 10 times or more) as large as the number of
variables in the study.
4. For simple experimental research with tight experimental controls (matched pairs,
etc.), successful research is possible with samples as small as 10 to 20 in size.
KEY TERMS
1. "What is Sample Design"? What all are the points to be considered to develop a
sample design?
2. Explain the various sampling methods under probability sampling.
3. Discuss the non probability sampling methods.
4. What are the importance of sample size and sampling design?
5. Discuss the other sampling methods.
6. Explain why cluster sampling is a probability sampling design.
7. What are the advantages and disadvantages of cluster sampling?
8. Explain what precision and confidence are and how they influence sample size.
9. The use of a convenience sample used in organizational research is correct because all
members share the same organizational stimuli and go through almost the same kinds
of experience in their organizational life. Comment.
10. Use of a sample of 5,000 is not necessarily better than one of 500. How would you
react to this, statement?
11. Non-probability sampling designs ought to be preferred to probability sampling
designs in some cases. Explain with an example
OBJECTIVES
STRUCTURE
Interviewing
Training for interviewing
The major principle for asking questions
Probing
Recording the response
INTRODUCTION
A personal interviewer administering a questionnaire door to door, a telephone
interviewer calling from a central location, an observer counting pedestrians in a
shopping mall, and others involved in the collection of data and the supervision of that
process are all Field Workers in the field. The activities of the field workers may vary in
nature. This lesson would help to understand the interview methods in data collection
process of the research and field work management.
Who conducts the field work?
Data collection is rarely carried out by the person who designs the research project.
However, the data collecting stage is crucial, because the research project is no better
than the data collected in the field. Therefore, it is important that the research
administrator select capable people who may be entrusted to collect the data. An irony
of business research is that highly educated and trained individuals design the research,
but the people who collect the data typically have little research training or experience.
Knowing the vital importance of data collected in the field, research administrators
must concentrate on carefully selecting field workers.
INTERVIEWING
Interviewing process is establishing rapport with the, respondent Interviewer bias may
enter in if the field worker's clothing or physical appearance is unattractive or unusual.
Suppose that a male interviewer, wearing a dirty T-shirt, interviews subjects in an
upper-income neighborhood. Respondents may consider the interviewer slovenly and
be less cooperative than they would be with a person dressed more appropriately.
Interviewers and other fieldworkers are generally paid an hourly rate or a per-interview
fee. Often interviewers are part-time workers, housewives, graduate students, secondary
school teachers from diverse backgrounds. Primary and secondary school teachers are
an excellent source for temporary interviewers during the summer, especially when they
conduct interviews outside the school districts where they teach. Teachers' educational
backgrounds and experiences with the public make them excellent candidates for
fieldwork.
TRAINING FOR INTERVIEWERS
The objective of training is to ensure that the data collection instrument is administered
uniformly by all field investigators. The goal of training sessions is to ensure that each
respondent is provided with common information. If the data are collected in a uniform
manner from all respondents, the training session will have been a success. After
personnel are recruited and selected, they must be trained.
Example: A woman who has just sent her youngest child off to first grade is hired by an
interviewing firm. She has decided to become a working mother by becoming a
professional interviewer. The training that she will receive after being selected by a
company may vary from virtually no training to a 3-day program if she is selected by one
of the larger survey research agencies. Almost always there will be a briefing session on
the particular project. Typically, the recruits will record answers on a practice
questionnaire during a simulated training interview.
More extensive training programs are likely to cover the following topics:
1. How to make initial contact with the respondent and secure the interview
2. How to ask survey questions
3. How to probe
4. How to record responses
5. How to terminate the interview
Making Initial Contact and Securing the Interview
Interviewers are trained to make appropriate opening remarks that will convince the
person that his or her cooperation is important.
Example: "Good afternoon, my name is _____ and I'm from a national survey
research company. We are conducting a survey concerning. I would like to get a few of
your ideas".
Much fieldwork is conducted by research suppliers who specialize in data collection.
When a second party is employed, the job of the study designed by the parent firm is not
only to hire a research supplier but also to establish supervisory controls over the field
service.
In some cases a third party is employed. For example, a firm may contact a survey
research firm, which in turn subcontracts the fieldwork to a field se vice. Under these
circumstances it is still desirable to know the problems that might occur in the field and
the managerial practices that can minimize them.
Asking the Questions
The purpose of the interview is, of course, to have the interviewer ask questions and
record the respondent's answers. Training in the art of stating questions can be
extremely beneficial, because interviewer bias can be a source of considerable error in
survey research.
There are five major principles for asking questions:
i. Ask the questions exactly as they are worded in the questionnaire.
ii. Read each question very slowly.
iii. Ask the questions in the order in which they are presented in the questionnaire.
iv. Ask every question specified in the questionnaire.
v. Repeat questions those are misunderstood or misinterpreted.
Although interviewers are generally trained in these procedures, when working in the
field many interviewers do not follow them exactly. Inexperienced interviewers may not
understand the importance of strict adherence to the instructions. Even professional
interviewers take shortcuts when the task becomes monotonous. Interviewers may
shorten questions or rephrase unconsciously when they rely on their memory of the
question rather than reading the question as it is worded. Even the slightest change in
wording can distort the meaning of the question and cause some bias to enter into a
study. By reading the question, the interviewer may be reminded to concentrate on
avoiding slight variations in tone of voice on particular words phases in the question.
PROBING
General training of interviewers should include instructions on how to probe when
respondents give no answer, incomplete answers, or answers that require clarification.
Probing may be needed for two types of situations. First, it is necessary when the
respondent must be motivated to enlarge on, clarify or explain his or her answer. It is
the interviewer's job to probe for complete, unambiguous answers. The interviewer must
encourage the respondent to clarify or expand on answers by providing a stimulus that
will not suggest the interviewer's own ideas or attitudes. The ability to probe with
neutral stimuli is the mark of an experienced interviewer. Second, probing may be
necessary in situations in which the respondent begins to ramble or lose track of the
question. In such cases the respondent must be led to focus on the specific content of
the interview and to avoid irrelevant and unnecessary information.
The interviewer has several possible probing tactics to choose from,
depending on the situation:
i. Repetition of the question: The respondent who remains completely silent may
not have understood the question or may not have decided how to answer it. Mere
repetition may encourage the respondent to answer in such cases. For example, if the
question is "What is there that you do not like about your supervisor?" and the
respondent does not answer, the interviewer may probe: "Just to check; is there
anything you do not like about your supervisor?"
ii. An expectant pause: If the interviewer believes the respondent has more to say,
the "silent probe" accompanied by an expectant look, may motivate the respondent to
gather his or her thoughts and give a complete response Of course, the interviewer must
be sensitive to the respondent so that the silent probe does not become an embarrassed
silence.
iii. Repetition of the Respondent's Reply: Sometimes the interviewer may repeat
the verbatim of the respondent. This may help the respondent to expand the answer.
RECORDING THE RESPONSES
The analyst who fails to instruct fieldworkers in the techniques of recording answers for
one study rarely forgets to do so in the second study. Although the concept of recording
an answer seems extremely simple, mistakes can be made in the recording phase of the
research. All fieldworkers should use the same mechanics of recording.
Example: It may appear insignificant to the interviewer whether she uses a pen or
pencil, but to the editor who must erase and rewrite illegible words, using a pencil is
extremely important.
The rules for recording responses to closed questionnaires vary with the specific
questionnaire. The general rule, however, is to place a check in the box that correctly
reflects the respondent's answer. All too often interviewers don't bother recording the
answer to a filter question because they believe that the subsequent answer will make
the answer to the filter question obvious. However, editors and coders do not know how
the respondent actually answered a question.
Managers of the field operation select, train, supervise, and control fieldworkers. Our
discussion of fieldwork principles mentioned selection and training. This section
investigates the tasks of the fieldwork manager in greater detail.
Briefing Session for Experienced Interviewers
After interviewers have been trained in fundamentals, and even when they have become
experienced, it is always necessary to inform workers about the individual project. Both
experienced and inexperienced fieldworkers must be instructed on the background of
the sponsoring organization, sampling techniques, asking questions, callback
procedures, and other matters specific to the project.
If there are special instructions - for example, about using show cards or video
equipment or about restricted interviewing times-these should also be covered during
the briefing session. Instructions for handling certain key questions are always
important. For example, the following fieldworker instructions appeared in a survey of
institutional investors who make buy-and sell decisions about stocks for banks, pension
funds, and the like.
A briefing session for experienced interviewers might go like, All interviewers report to
the central office, where the background of the firm and the general aims of the study
are briefly explained. Interviewers are not provided with too much information about
the purpose of the study, thus ensuring that they will not transmit any preconceived
notions to respondents. For example, in a survey about the banks in a community, the
interviewers would be told that the research is a banking study, but not the name of the
sponsoring bank. To train the interviewers about the questionnaire, a field supervisor
conducts an interview with another field supervisor who acts as a respondent. The
trainees observe the interviewing process, after which they each interview and record
the responses of another field supervisor. Additional instructions are given to the
trainees after the practice interview.
Training to Avoid Procedural Errors in Sample Selection
The briefing session also covers the sampling procedure. A number of research projects
allow the interviewer to be at least partially responsible for selection of the sample.
When the fieldworker has some discretion in the selection of respondents, the potential
for selection bias exists. This is obvious in the case of quota sampling, but less obvious
in other cases.
Example: In probability sampling where every nth house is selected, the fieldworker
uses his or her discretion in identifying housing units. Avoiding selection error may not
be as simple as it sounds.
Example: In an older, exclusive neighborhood, a mansion's coach house or servant's
quarters may have been converted into an apartment that should be identified as a
housing unit. This type of dwelling and other unusual housing units (apartments with
alley entrances only, lake cottages, rooming houses) may be overlooked, giving rise to
selection error. Errors may also occur in the selection of random digit dialing samples.
Considerable effort should be expended in training and supervisory control to minimize
these errors.
The activities involved in collecting data in the field may be performed by the
organization needing information, by research suppliers, or by third party field service
organizations. Proper execution of fieldwork is essential for producing research results
without substantial error.
Proper control of fieldwork begins with interviewer selection. Fieldworkers should
generally be healthy, outgoing, and well groomed. New fieldworkers must be trained in
opening the interview, asking the questions, probing for additional information,
recording the responses, and terminating the interview. Experienced fieldworkers are
briefed for each new project so that they are familiar with its specific requirements. A
particular concern of the briefing session is reminding fieldworkers to adhere closely to
the prescribed sampling procedures.
Careful supervision of fieldworkers is also necessary. Supervisors gather and edit
questionnaires each day. They check to see that field procedures are properly followed
and that interviews are on schedule. They also check to be sure that the proper sampling
units are used and that the proper people are responding in the study. Finally,
supervisors check for interviewer cheating and verify a portion of the interviews by reinterviewing a certain percentage of each fieldworker's respondents.
SUMMARY
This paper outlined the importance of training for new interviewers. In this chapter five
major principles for asking questions have been dealt in detail.
KEY TERMS
Field worker
Probing
Field interviewing
Briefing session
Training Interview
Reinterviewing questions
QUESTIONS
1. What qualities should a field worker possess?
2. What is the proper method of asking questions?
3. When should an interviewer probe? Give examples of how probing should be done?
4. How should an Interviewer terminate the interview?
5. What are the qualities of the interviewer that make him more effective?
REFERENCES
1. Ramanuj Majumdar, Marketing Research, Wiley Eastern Limited New Delhi (1991)
2. Cochran, W.G., Sampling Techniques, 2nd ed. New York: John Wiley and Sons.
3. Chaturvedi, J.C., Mathematical Statistics, Agra: Nok Jhonk Karyalaya, 1953.
STRUCTURE
INTRODUCTION
The availability of data source is very much needed to solve the problem and there are
many ways by which the data is collected. The task of data collection begins after a
research problem has been defined and research designed is prepared. Thus the data to
be collected can be classified as being either secondary or primary. The determination of
data source is based on three fundamental dimensions as given below:
turn involves travel plan need to verify and finally data to be tabulated. All these need
large funds which can be utilized elsewhere if secondary data can serve the purpose.
2. Another advantage is that the use of secondary data saves much of the time of the
researcher. This also leads to prompt completion of research project.
3. Secondary data are helpful not only because it is useful but the familiarity with the
data indicates deficiencies and gaps. As a result, the researcher can make the primary
data collection more specific and more relevant to the study.
4. It also helps in gaining new insights to the problem, then can be used to fine tune the
research hypothesis and objectives.
5. Finally, secondary data can be used as a basis of comparison with primary data that
has been collected for this study.
DISADVANTAGES OF SECONDARY DATA
An inherent disadvantage of secondary data is that it is not designed specifically to meet
the researcher's needs. Secondary data quickly becomes outdated in our rapidly
changing environments. Since the purpose of the most of the studies is to predict the
future, the secondary data must be timely. Hence the most common problems with
secondary data are:
1. Outdated information
2. Variation in definition of terms or classifications. The unit of measurement may cause
problems if they are not identical to the researcher's needs. Even though original units
were comparable, the aggregated or adjusted units of measurements are not suitable for
the present study. When the data are reported in a format that does not exactly meet the
researchers needs, the data conversion may be necessary.
3. Another disadvantage of secondary data is that the user has to control over their
accuracy even though it is timely & pertinent, they may be inaccurate.
THE NATURE AND SCOPE OF SECONDARY DATA
Focusing on the particular business or management problem, the researcher needs to
determine whether useful information already exists, of exists how relevant the
information. Since existing information are more widespread than one might expect.
The secondary data exists in three forms:
1. Internal secondary data: The data collected by the individual company for some
purpose and reported periodically. This is also called as primary sources. The primary
sources are original work of research or raw data without interpretation or
pronouncements that represent an official opinion or position - Memos, complete
interviews, speeches, laws, regulations, court decisions, standards, and most
government data, including census, economic and labor data. Primary sources are
always the most authoritative because the information is not filtered. It also includes
inventory records, personnel records, process charts and similar data.
2. External secondary data: It consists of data collected by outside agencies such as
government, trade associations of periodicals. This is called as secondary sources.
Encyclopedia, text books, handbooks, magazine and newspaper articles and most news
crafts are considered to secondary sources. Indeed all reference materials fall into this
category.
3. Computerized data sources: This includes internal and external data usually
collected by specific companies with online information sources. This can be called as
territory sources. These are represented by indexes, bibliography and other finding aids,
e.g., internet search engines.
Evaluation of secondary data sources
The emphasis on secondary data will increase if an attempt is made to establish a set of
procedures to evaluate the secondary data regarding the quality of information obtained
via secondary data sources. Specifically, if secondary data are to be used to assist in the
decision process, then they should be assessed according to the following principles.
1. Purpose: Since most secondary data are collected for purpose other than the one at
hand, the data must be carefully evaluated on how they relate to the current research
objectives. Many times the original collection of data is not consistent with the
particular research study. These inconsistencies usually result from the methods and
units of measurement.
2. Accuracy: When observing secondary data researchers need to keep in mind what
was actually measured. For example, if the actual purchases in a test market were
measured, did they measure the first-time trial purchases or repeat purchases?
Researchers must also asses the generality of the data.
3. Questions like i) was the data collected from certain groups only or randomly? ii)
was the measure developed properly? iii) was the data presented as the total of
responses from all respondents or were they categorized by age, sex or socio economic
status?
4. In addition to the above dimensions, researchers must assess when the data were
collected. This factor not only damages the accuracy of the data but also may be useless
for interpretation. Researchers also must keep in mind that the flaws in the research
design and methods will alter the current research in process.
5. Consistency: When evaluating any source of secondary data, a good strategy is to
seek out multiple sources of the same data to assure consistency. For example, when
evaluating the economic characteristics of a foreign market, a researcher may try to
gather the same information from government sources, private business publications
and specially import or export trade publications.
6. Credibility: Researcher should always question the credibility of the secondary data
source. Technical competence, service quality, reputation, and tracing and expertise of
personnel representing the organization are some measures of credibility.
7. Methodology: The quality of secondary data is only as good as the methodology
employed to gather them. Flaws in methodological procedures could produce results
that are invalid, unreliable or not generalizable beyond the study itself. Therefore,
researchers must evaluate the size and description of the sample, the response date, the
questionnaire, and the overall procedure for collecting the data (telephone, mail, or
personal interview).
8. Bias: Researchers must try to determine the underlying motivation or hidden
agenda, if any, behind the secondary data. It is not uncommon to find many secondary
data sources published to advance the interest of commercial, political or other intersect
groups. Researchers should try to determine if the organization reporting the report is
motivated by certain purpose.
A. Internal sources
Generally, internal data consists of sales or cost information. Data of this kind is found
in internal accounting or financial records. The two most useful sources of information
are sales invoices and accounts receivable reports; quarterly sales reports and sales
activity reports are also useful.
5. Supply characteristics
6. Regulations characteristics
7. International market characteristics
The external secondary data do not originate in the firm and are obtained from outside
sources. It may be noted that secondary data can be collected from the originating
sources or from secondary sources. For example, the office of economic advisor, GOI is
the originating source for wholesale prices. In contrast a publication such as RBI
bulletin on wholesale price is a secondary source.
These data may be available through Government publications, non-governmental
publications or syndicated services. Some examples are given below:
Government publications
1. Census by Registrar General of India
2. National Income by Central Statistical Organization also statistical abstract, annual
survey of industries.
3. Foreign trade by Director General of Commercial Intelligence.
4. Wholesale price index by Office of Economic Advisor
5. Economic Survey - Dept of Economic Affairs.
6. RBI Bulletin - RBI
7. Agricultural Situation in India - Ministry of Agriculture
8. Indian Labor Year Book - Labor Bureau
9. National Sample Survey Ministry of Planning
Non- government publications
Besides official agencies, there are number of private organizations which bring out
statistics in one form or another on a periodical basis of course industry and trade
associations are important like:
1. Indian Cotton Mills Federation or Confederation of Indian Textile Industry - about
textile industry.
2. Bombay Mill Owners Association - statistics of workers of mills.
Primary data
Secondary data
Advantages and disadvantages
Evaluation of secondary data
Sources of secondary data
Governmental publications
Syndicated services
QUESTIONS
1. Discuss the difference between primary and secondary data.
2. Explain the advantages and disadvantages of secondary data.
3. Write short notes on nature and scope of secondary data.
4. How will you evaluate the secondary data sources?
OBJECTIVES
STRUCTURE
Questionnaire design
Phrasing question
Art of asking question
Layout of traditional questionnaires
Many experts in survey research believe that improving the wording of questions can
contribute far more to accuracy than can improvements in sampling. Experiments have
shown that the range of error due to vague questions or use of imprecise words may be
as high as 20 or 30 percent. Consider the following example, which illustrates the
critical importance of selecting the word with the right meaning. The following
questions differ only in the use of the words should, could, and might:
Do you think anything should be done to make it easier for people to pay doctor
or hospital bills?
Do you think anything could be done to make it easier for people to pay doctor or
hospital bills?
Do you think anything might be done to make it easier for people to pay doctor or
hospital bills?
The results from the matched samples: 82 percent replied something should be done, 77
percent replied something could be done, and 63 percent replied something might be
done. Thus, a 19 percent difference occurred between the two extremes, should and
might. Ironically, this is the same percentage point error as in the Literary Digest Poll,
which is a frequently cited example or error associated with sampling.
The chapter outlines procedure for questionnaire design and illustrates that a little bit of
research knowledge can be a dangerous thing.
A Survey Is Only As Good As the Questions It Asks
Each stage of the business research process is important because of its interdependence
with other stages of the process. However, a survey is only as good as the questions it
asks. The importance of wording questions is easily overlooked, but questionnaire
design is one of the most critical stages in the survey research process.
"A good questionnaire appears as easy to compose as does a good poem. But it is usually
the result of long, painstaking word". Business people who are inexperienced in business
research frequently believe that constructing a questionnaire in a matter of hours.
Unfortunately, newcomers who naively believe that common sense and good grammar
are all that are needed to construct a questionnaire generally learn that their hasty
efforts are inadequate.
While common sense and good grammar are important in question writing, more is
required in the art of questionnaire design. To assume that people will understand the
questions is a common error. People simply may not know what is being asked. They
may be unaware of the product or topic interest, they may confuse the subject with
something else, or the question may not mean the same thing to everyone interviewed.
Respondents may refuse to answer personal questions. Further, properly wording the
questionnaire is crucial, as some problems may be minimized or avoided altogether if a
skilled researcher composes the questions.
QUESTIONNAIRE DESIGN: AN OVERVIEW OF THE MAJOR DECISIONS
Relevance and accuracy are the two basic criteria a questionnaire must meet if it is to
achieve the researcher's purpose. To achieve these ends, a researcher who systematically
plans a questionnaire's design will be required to make several decisions - typically, but
not necessarily, in the order listed below:
1. What should be asked?
2. How should each question be phrased?
3. In what sequence should the questions be arranged?
4. What questionnaire layout will best serve the research objectives?
5. How should the questionnaire be pretested? Does the questionnaire need to be
revised?
What Should Be Asked?
During the early stages of the research process, certain decisions will have been made
that will influence the questionnaire design. The preceding chapters stressed the need to
have a good problem definition and clear objectives for the study. The problem
definition will indicate which type of information must be collected to answer the
manager's questions; different types of questions may be better at obtaining certain
types of information than others. Further, the communication mediums used for data
collection - telephone interview, personal interview, or self-administered survey will
have been determined. This decision is another forward linkage that influences the
structure and content of the questionnaire. The specific questions to be asked will be a
function of the pervious decisions later stages of the research process also have an
important impact o questionnaire wording. For example, determination of the questions
that should be asked will be influenced by the requirements for data analysis. As the
questionnaire is being designed, the researcher should be thinking about the types of
statistical analysis that will be conducted.
Questionnaire Relevancy
A questionnaire is relevant if no unnecessary information is collected and if the
information that is needed to solve the business problem is obtained. Asking the wrong
or an irrelevant question is a pitfall to be avoided. If the task is to pinpoint
compensation problems, for example, questions asking for general information about
morale may be inappropriate. To ensure information relevancy, the researcher must be
specific about data needs, and there should be a rationale for each item of information.
After conducting surveys, many disappointed researchers have discovered that some
important questions were omitted. Thus, when planning the questionnaire design, it is
essential to think about possible omissions. Is information being collected on the
relevant demographic and psychographic variables? Are there any questions that might
clarify the answers to other questions? Will the results of the study provide the solution
to the manager's problem?
Questionnaire Accuracy
Once the researcher has decided what should be asked, the criterion of accuracy
becomes the primary concern. Accuracy means that the information is reliable and valid
while experienced researchers generally believe that one should use simple,
understandable, unbiased, unambiguous, nonirritating words, no step-by-step
procedure to ensure accuracy in question writing can be generalized across projects.
Obtaining accurate answers from respondents is strongly influenced by the researcher's
ability to design a questionnaire that facilitates recall and that will motivate the
respondent to cooperate.
Respondent tend to be most cooperative when, the subjects of the research is
interesting. Also, if questions are not lengthy, difficult to answer, or ego threatening,
there is higher probability of obtaining unbiased answers, question wording and
sequence substantially influence accuracy. These topics are treated in subsequent
sections of this chapter.
PHRASING QUESTIONS
There are many ways to phrase question, and many standard question formats have
been developed in previous research studies. This section presents h classification of
question types and provides some helpful guidelines to researchers who must write
questions.
Open-Ended Response versus Fixed-Alternative Questions
Questions may be categorized as either of two basic types, according to the amount of
freedom respondents are given in answering them. Response questions pose some
problem or topic and ask the respondent to answer in his or her own words. For
example:
What things do you like most about your job?
What names of local banks can you think of offhand?
What comes to mind when you look at this advertisement?
Do you think that there are some ways in which life in the United States is getting
worse? How is that?
If the question is asked in a personal interview, the interviewer may probe for more
information by asking such questions as: Anything else? or Could you tell me more
about your thinking on that? Open-ended response questions are free-answer
questions. They may be contrasted to the fixed-alternative question, sometimes called a
"closed question", in which the respondent is given specific, limited-alternative
responses and asked to choose the one closest to his or her own viewpoint. For example:
Did you work overtime or at more than one job last week?
Yes ____ No _____
Compared to ten years ago, would you say that the quality of most products made in
Japan is higher, about the same, or not as good?
Higher ____ About the same _____ Not as good _____
Open-ended response questions are most beneficial when the researcher is conducting
exploratory research, especially if the range of responses is not known. Open-ended
questions can be used to learn what words and phrases people spontaneously give to the
free-response questions. Respondents are free to answer with whatever is uppermost in
their thinking. By gaining free and uninhibited responses, a researcher may find some
unanticipated reaction toward the topic. As the responses have the "flavor" of the
conversational language that people use in talking about products or jobs, responses to
these questions may be a source for effective communication.
No _____
Several types of questions provide the respondent with multiple-choice alternatives. The
determinant-choice questions require the respondent to choose one and only one
response from among several possible alternatives. For example:
Please give us some information about your flight. In which section of the aircraft did
you sit?
First Class _______ Business Class ______
health threat to you personally is the inhaling of this secondhand smoke, often called
passive smoking: Is it a very serious health threat, somewhat serious, not too serious,
or not serious at all?
1. Very serious
2. Somewhat serious
3. Not too serious
4. Not serious at all
5. Don't know
THE ART OF ASKING QUESTIONS
In develop a questionnaire, there are no hard-and-fast rules. Fortunately, however,
some guidelines that help to prevent the most common mistakes have "been developed
from research experience.
1. Avoid Complexity: Use Simple, Conversational Language
Words used in questionnaires should be readily understandable to all respondent. The
researcher usually has the difficult task of adopting the conversational language of
people from the lower educational levels without talking down to better-educated
respondents. Remember, not all people have the vocabulary of a college student. A
substantial number of Americans never go beyond high school.
Respondents can probably tell an interviewer whether they are married, single,
divorced, separated, or windowed, but providing their "marital status" may present a
problem. Also, the technical jargon of corporate executives should be avoided when
surveying retailers, factory employees, or industrial users. "Marginal analysis,"
"decision support systems," and other words from the language of the corporate staff
will not have the same meaning to- or be understood by- a store owner / operator in a
retail survey. The vocabulary in following question (from an attitude survey on social
problems) is probably confusing for many respondents:
When effluents from a paper mill can be drunk, and exhaust from factory smokestacks
can be breathed, then humankind will have done a good job in saving the
environment... Don't you agree that what we want is zero toxicity and no effluents?
This lengthy question is also a leading question.
2. Avoid Leading and Loaded Questions
Leading and loaded questions are a major source of bias in question wording. Leading
Questions suggest or imply certain answers. In a study of the dry-cleaning industry, this
question was asked:
Many people are using dry cleaning less because of improved wash-and-wear clothes.
How do you feel wash-and-wear clothes have affected your use of dry-cleaning
facilities in the past 4 years?
_______Use less
______No change
__________Use more
The potential "bandwagon effect" implied in this question threatens the study's validity.
Loaded questions suggest a socially desirable answer or are emotionally charged.
Consider the following:
In light of todays farm crisis, it would be in the public's best interest to have the
federal government require labeling of imported meat.
_____Strongly ____Agree ____Uncertain
Disagree
____Strongly disagree
____
Answers might be different if the loaded portion of the statement, "farm crisis" had
another wording suggesting a problem of less magnitude than a crisis. A television
station produced the following 10-second spot asking for viewer feedback:
We are happy when you like programs on Channel 7. We are sad when you dislike
programs on Channel 7. Write to us and let us know what you think of our
programming.
Most people do not wish to make others sad. This question is likely to elicit only positive
comments. Some answers to certain questions are more socially desirable than others.
For example, a truthful answer to the following classification question might be painful
Where did you rank academically in your high school graduating class?
___Top quarter ___2nd quarter
___3rd quarter
___4th quarter
When taking personality tests, respondents frequently are able to determine which
answers are most socially acceptable, even though those answers do not portray their
true feelings.
3. Avoid Ambiguity: Be a Specific as Possible
Items on questionnaires are often ambiguous because they are too general. Consider
indefinite words such as often, usually, regularly, frequently, many, good, fair, and poor.
Each of these words has many meanings. For one person, frequent reading of Fortune
magazine may be reading six or seven issues a year; for another it may be two issues a
year. The word fair has a great variety of meanings; the same is true for many indefinite
words.
Questions such as the following should be interpreted with care:
How often do you feel that you can consider all of the alternatives before making a
decision to follow a specific course of action?
___Always
___Fairly
___Occasionally
___ No
The question contains the implicit assumption that people believe the gift-wrapping
program is excellent. By answering yes, the respondent implies that the program is, in
fact, excellent and that things are just fine as they are. By answering no, he or she
implies that the store should discontinue the gift wrapping. The researcher should not
place the respondent in that sort of bind by including an implicit assumption in the
question.
6. Avoid Burdensome Questions That May Tax the Respondent's Memory
A simple fact of human life is that people forget. Researchers writing questions about
past behavior or events should recognize that certain questions may make serious
demands on the respondent's memory. Writing questions about prior events requires a
conscientious attempt to minimize the problem associated with forgetting.
It many situations, respondents cannot recall the answer to a question. For example, a
telephone survey conducted during the 24-hour period following airing of the Super
Bowl might establish whether the respondent watched the Super Bowl and then ask: "Do
you recall any commercials on that program?" If the answer is positive, the interviewer
might ask: "what brands were advertised?" These two questions measure unaided
recall, because they give the respondent no clue as to the brand of interest.
What is the Best Question Sequence?
The order of questions, or the question sequence, may serve several functions for the
researcher. If the opening questions are interesting, simple to comprehend, and easy to
answer, respondents' cooperation and involvement can be maintained throughout the
questionnaire. Asking easy-to-answer questions teaches respondents their role and
builds confidence; they know this is a researcher and not another salesperson posing as
an interviewer. If respondents' curiosity is not aroused at the outset, they can become
disinterested and terminate the interviewer. A mail research expert reports that a mail
survey terminates the interview. A mail research expert reports that a mail survey
among department store buyers drew an extremely poor return. However, when some
introductory questions related to the advisability of congressional action on pending
legislation of great importance to these buyers were placed first on the questionnaire, a
substantial improvement in response rate occurred. Respondents completed all thequestions, not only those in the opening section.
In their attempts to "warm up" respondents toward the questionnaire dependent
researchers frequently ask demographic or classification questions at the beginning of
the questionnaire. This is generally not advisable. It may embarrass or threaten
respondents. It is generally better to ask embarrassing questions at the middle or end of
the questionnaire, after rapport has been established between respondent and
interviewer.
Sequencing specific questions before asking about broader issues is a common cause of
order bias. For example, bias may arise if questions about a specific clothing store are
asked prior to those concerning the general criteria for selecting a clothing store.
Suppose a respondent who indicates in the first portion of a questionnaire that the
shops at a store where parking needs to be improved. Later in the questionnaire, to
avoid appearing inconsistent, she may state that parking is less important a factor than
she really believes it is. Specific questions may thus influence the more general ones.
Therefore, it is advisable to ask general questions before specific questions to obtain the
freest of open ended responses. This procedure, known as the funnel technique, allows
for researcher to understand the respondents frame of reference before asking more
specific questions about the level of the respondents information and the intensity of
his or her opinions.
One advantage of internet surveys is the ability to reduce order bias by having the
computer randomly order questions and/or response alternatives. With complete
randomization, question order is random and respondents see response alternatives in
random positions. Asking a question that doesnt apply to the respondent or that the
respondent is not qualified to answer may be irritating or may cause a biased response
because the respondent wishes to please the interviewer or to avoid embarrassment.
Including a filter question minimizes the chance of asking questions that are
inapplicable. Asking "where do you generally have cheque-cashing problems in Delhi"
may elicit a response even though the respondent has not had any cheque-cashing
problems and may simply wish to please the interviewer with an answer. A filter
question such as:
Do you ever have a problem cashing a cheque in Delhi? ___Yes
___No
would screen out the people who are not qualified to answer.
Another form of filter question, the pivot question, can be used to obtain income
information and other data that respondents may be reluctant to provide. For example,
a respondent is asked.
"Is your total family income over Rs.75,000?" IF NO, ASK...
"Is it over or under Rs.50,000?" IF UNDER, ASK
"Is it over or under Rs.25,000?"
So, the options are
1. Over Rs.75,000
2. Rs.50,001 - Rs.75,000
3. Rs.25,001 - Rs.50,000
4. Under Rs.25,001
Structuring the order of questions so that they are logical will help to ensure the
respondents cooperation and eliminate confusion or indecision. The researcher
maintains legitimacy by making sure that the respondent can comprehend the
relationship between a given question and the overall purpose of the study.
What is the best layout?
Good layout and physical attractiveness are crucial in mail, Internet, and other selfadministered questionnaires. For different reasons it is also important to have a good
layout in questionnaires designed for personal and telephone interviews.
LAYOUT OF TRADITIONAL QUESTIONNAIRES
The layout should be neat and attractive, and the instructions for the interviewer should
be clear. Often money can be spent on an incentive to improve the attractiveness and
quality of the questionnaire.
Mail questionnaires should never be overcrowded. Margins should be of decent size,
white space should be used to separate blocks of print, and any unavoidable columns of
multiple boxed should be kept to a minimum. All boldface capital letters should easy to
follow.
Questionnaires should be designed to appear as brief and small as possible. Sometimes
it is advisable to use a booklet form of questionnaire, rather than a large number of
pages stapled together.
In situations where it is necessary to conserve space on the questionnaire or to facilitate
entering the data into a computer or tabulating the data, a multiple-grid layout may be
used. In this type of layout, a question is followed by corresponding response
alternatives arranged in a grid or matrix format.
Experienced researchers have found that is pays to phrase the title of the questionnaire
carefully. In self-administered and mail questionnaires a carefully constructed title may
by itself capture the respondents interest, underline the important of the research
("Nationwide Study of Blood donors"), emphasize the interesting nature of the study
("Study of Internet Usage"), appeal to the respondents ego ("Survey among Top
Executives"), or emphasize the confidential nature of the study ("A Confidential Survey
among"). The researcher should take steps to ensure that the wording of the title will
not bias the respondent in the same way that a leading question might.
When an interviewer is to administer the questionnaire, the analyst can design the
questionnaire to make the job of following interconnected questions much easier by
utilizing instruction, directional arrows, special question formats, and other tricks of the
trade.
SUMMARY
Many novelists write, rewrite, and revise certain chapters, paragraphs, and even
sentences of their books. The research analyst lives in a similar world. Rarely does one
write only a first draft of a questionnaire. Usually, the questionnaire is tried out on a
group that is selected on a convenience basis and that is similar in makeup to the one
that ultimately will be sampled. Researchers should select a group that is not too
divergent from the actual respondents. (e.g. business students as surrogates for business
people), but it is not necessary to get a statistical sample for protesting. The protesting
process allows the researchers to determine if the respondents have any difficulty
understanding the questionnaire and whether there are any ambiguous or biased
questions. This process is exceedingly beneficial. Making a mistake with 25 or 50
subjects can avert the disaster of administering an invalid questionnaire to several
hundred individuals.
KEY TERMS
QUESTIONS
1. What is the difference between leading question and loaded question?
2. Design an open end question to measure a reaction to a particular advertisement.
3. Design a complete a questionnaire to evaluate job satisfaction.
4. Develop a checklist to consider in questionnaire construction.
MEASUREMENT
OBJECTIVES
STRUCTURE
Measurement
Scale measurement
Types of scale
Criteria for good measurement
MEASUREMENT
Measurement is an integral part of the modern world. Today we have progressed in the
physical sciences to such as extent that we are now able to measure the rotation of a
distant star, the attitude in micro-inches and so on. Today such a precise physical
measurement is very critical. In many business situations, the majority of the
measurements are applies to things that are much more abstract than attitude or time.
The accurate measurement is essential for effective decision making. The purpose of this
chapter is to provide with a basic understanding of the measurement process and rules
needed for developing sound scale measurements.
In management research, measurement is viewed as the integrative process of
determining the amount (intensity) of information about constructs, concepts or objects
of interest and their relationship to a defined problem or opportunity. It is important to
understand the two aspects of measurement one is construct development, which
provides necessary and precise definition which begins the research process called
problem definition in turn determine what specific data should be collected. Another is
scale measurement means how the information is collected with reference to construct.
In other words, the goal of construct development is to precisely identify and define
what is to be measured including dimensions. In turn, the goal of scale measurement is
to determine how to precisely measure the constructs.
Regardless of whether the researcher is attempting to collect primary data or secondary
data, all data can be logically classified as under.
a) State-of-Being Data
When the problem requires collecting responses that are pertinent to the physical,
demographical or socioeconomic characteristics of individuals, objects or organizations,
the resulting raw data are considered as state-of-being data. This data represent factual
characteristics that can be verified through several sources other than the persons
providing the information.
b) State-of-Mind Data
This represents that mental attributes of individuals that are not directly observable or
available through some other external sources. It exists only within the minds of people.
The researcher has to ask a person to respond the stated questions. Examples are
personality traits, attitudes, feelings, perceptions, beliefs, awareness level, preferences,
images etc.
c) State-of-Behavior Data
This represents an individuals or organizations current observable actions or reactions
or recorded past actions. A person may categorically ask the past behavior. This can be
checked using external secondary sources, but that is very difficult process in terms of
time, effort and accuracy.
d) State-of-Intension Data
This represents individuals or organizations expressed plans of future behavior. Again
this also collected by asking carefully defined questions. Like the above data, this also
very difficult to verify through external, secondary sources, but verification is possible.
With the background information about the type of data which are collected, the
following pages will be very useful in understanding the concepts of scale measurement.
SCALE MEASUREMENT
Scale measurement can be defined as the process of assigning a set of descriptions to
represent the range of possible responses to a question about a particular object or
construct. Scale measurement directly determines the amount of raw data that can be
ascertained from a given questioning or observation method. This attempts to assign
designated degrees of intensity to the responses, which are commonly referred to as
scale points. The researcher can control the amount of raw data that can be obtained
from asking questions by incorporating scale properties or assumption in scale points.
There are four scaling properties that a researcher can use in developing scales namely
assignment, order distance and origin.
1. Assignment (also referred to as description or category property): It is the
researchers employment of unique description to identity each object within a set, e.g.,
the use of numbers, colors, yes & no responses.
2. Order refers to the relative magnitude between the raw responses. It establishes and
creates hierarchical rank-order relationship coming objects, e.g., 1st place is better than
4th place.
3. Distance, is the measurement the express the exact difference between the two
responses. This allows the researcher and respondent to identify, understand, and
accurately express absolute difference between objects, e.g., Family A has 3 children and
Family B has 6 children.
4. Origin, refers to the use of a unique starting as being "true zero" e.g., asking
respondent his or her weight or current age, market share of specific brand.
TYPES OF SCALES
While scaling properties determine the amount of raw data that can be obtained from
any scale design, all questions and the scale measurement can be logically and
accurately classified as one of four basic scale types: nominal, ordinal, integral or ratio.
A scale may be defined as any series of items that are arranged progressively according
to value or magnitude, into which an item can be placed according to its quantification.
The following table represents the relationship between types of scales & scaling
properties:
1. Nominal Scale
In business research, nominal data are probably more widely collected than any other. It
is the simplest type of scale and also the most basic of the four types of scale designs. In
such a scale, the numbers serve as labels to identify persons, objects or events. Nominal
scales are the least powerful of the four data types. They suggest no order or distance
relationship and have no arithmetic origin. This scale allows the researcher only to
categorize the raw responses into mutually exclusive and collectively exhaustive. In the
nominal scale, the only operation is the counting of numbers in each group. An example
of typical nominal scale in business research is the coding of males as 1 and females as 2.
Example 1:
Please indicate your current marital status.
____Married
____Single
Example 2:
How do you classify yourself?
____Indian ___American ___Asian ___Black
2. Ordinal Scale
As the name implies, these are ranking scales. Ordinal data include the characteristics of
the nominal data plus an indicator of order, means, this data activates both are
assignment and order scaling properties. The researcher can rank-order the raw
responses into a hierarchical pattern. The use of ordinal data scale implies a statement
of "greater than" or "less than" without stating how much greater or less. Examples of
ordinal data include opinion and preference scales. A typical ordinal scale in business
research asks respondents to rate career opportunities, brands, companies etc., as
excellent, good, fair or poor.
Example 1:
Which of the following one category best describes your knowledge about computers?
1) Excellent 2) Good
3) Basic
4) Little
5) No knowledge
Example 2:
Among the listed below, please indicate top three preference using 1, 2, 3 as your
choice in the respective source provided:
By post
By courier
By telephone
By speed post
By internet
By person
Also to be noted is that the individual ranking can be combined and get a collective
ranking of a group.
3. Interval Scales
The structure of this scale not only show the assignment, order scaling properties but
also the distance property with interval scale, researchers can identity not only some
type of hierarchical order among to raw data but also the specific differences between
the data. The classic example of this scale is the Fahrenheit temperature scale. If a
temperature is 80 degree, it cannot be said that it is twice as hot as 40 degree. The
reason is that 0 degree does not represent the lack of temperature, but a relative point
on the Fahrenheit scale. Similarly, when this scale is used to measure psychological
attributes, the researcher can comment on the magnitude of differences or compare the
average differences but cannot determine the actual strength of attitude toward an
object. However many attitude scales are presumed to be interval scales. Interval scales
are more powerful than nominal and ordinal scales. Also they are quicker to complete
and it is convenient for researcher.
Example 1:
Into which of the following categories does your income fall?
1. Below 5000
2. 5000 - 10,000
3. 10,000 - 15,000
4. 15,000 - 25,000
5. above 25,000
Example 2:
Approximately how long you lived in the current address?
1. Less than 1 year
2. 1-3 years
3. 4-6 years
4. More than 6 years
4. Ratio Scales
This is the only scale that simultaneously activates all four scaling properties. A ratio
scale tends to be the most sophisticated scale in the sense that it allows not only to
identify the absolute differences between each represents but also absolute
comparisons.
Examples of ratio scales are the commonly used physical dimensions such as height,
weight, distance, money value and population counts. It is necessary to remember that
ratio scale structures are designed to allow a "zero" or "true state of nothing" response to
be a valid raw response to the question. Normally, the ratio scale requests that
respondents give a specific singular numerical value as their response, regardless of
whether or not a set of scale points used. The following are the examples of ratio scales:
Example 1:
Please circle the numbers of children below 18 years of ages in your house?
0
Example 2:
In past seven days, how many times did you go to retail shop?
____ number of times
Mathematical and Statistical Analysis of Scales
The type of scale that is utilized in business research will determine the form of
statistical analysis. For example certain operations can be conducted only if a scale, of a
particular nature. The following will show the relationship between scale types and
measures of central tendency and dispersion.
Here,
A - Appropriate
More A - More appropriate
Most A - Most appropriate
IA - Inappropriate
Criteria for good measurement
There are four major criteria for evaluating measurement: reliability, validity, sensitivity
and practicality.
1. Reliability
It refers to the extent to which a scale can reproduce the same measurement results in
repeated trials. Reliability applies to a measure when similar results are obtained
overtime across situations. Broadly defined, reliability is the degree to which measures
are free from error and therefore yield consistent results. As discussed in the earlier
chapter the error in scale measurements leads to lower scale reliability. Two dimensions
underline the concept of reliability: one is repeatability and the other is internal
consistency.
First, the test-retest method involves administrating the same scale or measure to the
same respondents at two separate times to test for stability. If the measure is stable over
time, the test, administered under the same conditions each time, should obtain similar
results. The high stability correlation or consistency between the two measures at time 1
and 2 indicates a high degree of reliability.
The second dimension of reliability concerns the homogeneity of the measure. The
Split-half technique can be used when the measuring tool has many similar
questions or statements to which subjects can respond. The instrument is administered
and the results are separated by item into even and odd numbers or randomly selected
halves. When the two halves are correlated, if the result of the correlation is high, the
instrument is said to be high reliability in internal consistency.
The Spearman-Brown Correction Formula is used to adjust the effect of test
length and to estimate reliability of the whole set. But, this approach may influence the
integral consistency because of the way in which the test is split. In order to overcome
Kuder - Richardson Formula (KR 20) and Cronbach's Coefficient Alpha are
two frequently used examples. The KR 20 is the method from which alpha was
generalized and is used to estimate reliability for dichotomous items. Cronbach's alpha
has the most utility for multi-scale items at the interval level of measurement.
The third perspective on reliability considers how much error may be introduced by
different investigators or different samples of items being studied. In other words, the
researcher creates two similar yet different scale measurements for the given construct.
An example of this is the scoring of Olympic skaters by a panel of judges.
2. Validity
The purpose of measurement is to measure what it intend to measure; but this is not as
simple as it sounds at first. Validity is the ability of a measure to measure what it is
proposed to measure. If it does not measure what it is designated to measure, there will
be problems. To assess the validity there are second ways which are discussed here
under.
Face validity or Content validity, refers to the subjective agreement among
professionals that a scale logically appears to reflect accurately what it intend to
measure. When it appears evident to experts that the measure provides adequate
coverage of the concept, a measure has face validity.
Criterion - Related validity reflects the success of measures used for prediction or
estimation. Criterion validity may be classified as either concurrent validity or
predictive validity, depending on the time sequence in which the 'new' measurement
scale and the criterion measure are correlated. If the new measure is taken at the same
time as the criterion measure and shown to be valid, then it has concurrent validity.
Predictive validity is established when a new measure predicts a future event. These two
measures differ only on the basis of time.
Construct validity is established by the degree to which a measure confirms the
hypotheses generated from theory based on concepts. It implies the empirical evidence
generated by a measure with the theoretical logic. To achieve this validity, the
researcher may use convergent validity (should converge with similar measure) or
discriminant validity (when it has low correlation with the measures of dissimilar
concepts.)
3. Sensitivity
It is an important concept, particularly when changes in attitude or other hypothetical
constructs are under investigation. It refers to an instruments ability to accurately
measure variability in stimuli or responses. A dichotomous response category such as
"agree" or "disagree" does not allow attitude change. But the scale staring from "strongly
agree", "agree", "neither agree nor disagree", "disagree" and "strongly disagree"
increases the sensitivity.
4. Practicality
It can be defined in terms of economy, convenience and interpretability. This means the
scientific requirements for the measurement process is called reliable and valid, while
operational requirements called it as practical where the above mentioned three aspects
are more important.
SUMMARY
This paper outlined the importance of the measurement. Different types of scales have
been dealt in detail. This chapter has given the criteria for a good measurement.
KEY TERMS
Nominal scale
Ordinal scale
Integral scale
Ratio scale
Reliability
Split-half Technique
Spearman-Brown Correction Formula
Kuder - Richardson Formula (KR 20)
Cronbach's Coefficient Alpha
Validity
Face validity
Content validity
Criterion - Related validity - concurrent validity or predictive validity
Construct validity
Sensitivity
Practicality
QUESTIONS
1. What are different types of data in the attitude measurement could be collected?
2. Discuss the measurement scaling properties.
3. Explain different scales of measurement.
4. Is the statistical analysis is based on the type of scale? Explain.
5. What do you mean by good measurement?
6. Explain various methods of reliability and validity.
OBJECTIVES
STRUCTURE
ATTITUDE DEFINED
There are many definitions for the term attitude. An attitude is usually viewed as an
enduring disposition to respond consistently in a given manner to various aspects of the
world, including persons, events, and objects. One conception of attitude is reflected in
this brief statement: "Sally loves working at Sam's. She believes it's clean, conveniently
located, and has the best wages in town She intends to work there until she retires. In
this short description are three components of attitude: the affective, the cognitive,
and the behavioral.
The affective component reflects an individual's general feelings or emotion toward
an object. Statements such as "I love my job", "I liked that book, A Corporate Bestiary,
and "I hate apple juice" - reflect the emotional character of attitudes.
The way one feels about a product, a person, or an object is usually tied to ones beliefs
or cognitions. The cognitive component represents one's awareness of and
knowledge about an object. A woman might feel happy about her job because she
"believes that the pay is great" or because she knows "that my job is the biggest
challenge in India."
The third component of an attitude is the behavioral component. Intention and
behavioral expectations are included in this component, which therefore reflects a
predisposition to action.
Techniques for Measuring Attitudes
A remarkable variety of techniques have been devised to measure attitudes part, this
diversity stems from the lack, of consensus about the exact definite of the concept.
Further, the affective, cognitive, and behavioral component an attitude may be
measured by different means. For example, sympathetic nervous system responses may
be recorded using physiological measures to measure affect but they are not good
measures of behavioral intentions. Direct verbal statements concerning affect, belief, or
behavior are utilized to measure behavioral intent. However, attitudes may also be
measured indirectly by using the qualitative explanatory techniques. Obtaining verbal
statements from respondents generally requires that the respondent perform a task such
as ranking, rating, sorting, or making a choice or a comparison.
A ranking task requires that the respondents rank order a small number of items on
the basis of overall preference or some characteristic of the stimulus. Rating asks the
respondents to estimate the magnitude of a characteristic or quality that an object
possesses. Quantitative scores, along a continuum that has been supplied to the
respondents, are used to estimate the strength of the attitude or belief. In other words,
the respondents indicate the position, on a scale, where they would rate the object.
A sorting technique might present respondents with several product concepts,
printed on cards, and require that the respondents arrange the cards into a number of
piles or otherwise classify the product concepts. The choice technique, choosing one
of two or more alternatives, is another type of attitude measurement. If a respondent
chooses one object over another, the researcher can assume that the respondent prefers
the chosen object over the other.
The most popular techniques for measuring attitudes are presented in this chapter.
Physiological Measures of Attitudes
Measures of galvanic skin response, blood pressure, and pupil dilation and other
physiological measures may be utilized to assess the affective component of attitudes.
They provide a means of measuring attitudes without verbally questioning the
respondent. In general, they can provide a gross measure of like or dislike, but they are
not sensitive measures for identifying gradients of an attitude.
Attitude Rating Scales
Using rating scales to measure attitudes is perhaps the most common practice in
business research. This section discusses many rating, scales designed to enable
respondents to report the intensity of their attitudes.
Simple Attitude Scales
In this most basic form, attitude scaling requires that an individual agree or disagree
with a statement or respond to a single question. For example, respondents in a political
poll may be asked whether they agree or disagree with the statement "The president
should run for re-election", or an individual might be asked to indicate whether he likes
or dislikes labor unions. Because this type or self-rating scale merely classifies
respondents into one of two categories, it has only the properties of a nominal scale.
This, of course, limits the type of mathematical analysis that may be utilized with this
basic scale. Despite the disadvantages, simple attitude scaling may be used when
questionnaires are extremely long, when respondents have little education, or for other
specific reasons.
Most attitude theorists believe that attitudes vary along continua. An early attitude
researcher pioneered the view that the task of attitude scaling is to measure the distance
from "good" to "bad", "low" to "high", "like" to "dislike , and so on. Thus the purpose of
an attitude scale is to find an individual's position on the continuum. Simple scales do
not allow for making fine distinctions in attitudes. Several scales have been developed to
help make more precise measurements.
Category Scales
Some rating scales have only two response categories: agree and disagree. Expanding
the response categories provides the respondent more flexibility in the rating task. Even
more information is provided if the categories are ordered according to a descriptive or
evaluative dimension. Consider the questions below:
How often is your supervisor courteous and friendly to you?
___Never
Each of these category scales is a more sensitive measure than a scaled with only two
response categories. Each provides more information.
Wording is an extremely important factor in the usefulness of these scales. Exhibit 14.1
shows some common wordings for category scales.
Disagree
(2)
Uncertain
Agree
Strongly agree
(3)
(4)
(5)
2 Agree
3 Neutral
4 Disagree
5 - Strongly Disagree
Circle one and only one answer for each statement. There are no right or wrong
answers to these questions:
In Likert's original procedure, a large number of statements are generated and then an
item analysis is performed. The purpose of the item analysis is to ensure that final
items evoke a wide response and discriminate among those with positive and negative
attitudes. Items that are poor because they lack clarity or elicit mixed response patterns
are eliminated from the final statement list. However, many business researchers do not
follow the exact procedure prescribed by Likert. Hence, many business researches do
not follow the exact procedure prescribed by Likert. Hence, a disadvantage of the Likerttype summated rating method is that it is difficult to know what a single summated
score means. Many patterns of response to the various statements can produce the same
total score. Thus, identical total scores may reflect different "attitudes" because
respondents endorsed different combinations of statements.
Semantic Differential
The semantic differential is a series of attitude scales. This popular attitudemeasurement technique consists of presenting an identification of a company, product,
brand, job, or other concept, followed by a series of seven-point bipolar rating scales.
Bipolar adjectives, such as "good" and "bad", "modern" and "old-fashioned", or "clean"
and "dirty," anchor the beginning and end (or poles) of the scale.
Modern_____:______ :_____:______:_____:_____ :_____ Old-Fashioned
The subject makes repeated judgments of the concept under investigation on each of the
scales.
The scoring of the semantic differential can be illustrated by using the scale bounded by
the anchors "modern" and "old-fashioned." Respondents are instructed to check the
place that indicates the nearest appropriate adjective. From left to right, the scale
intervals are interpreted as extremely modern, very modern, slightly modern, both
modern and old-fashioned, slightly old-fashioned, very old-fashioned, and extremely
old-fashioned. A weight is assigned to each position on the rating scale. Traditionally,
scores are 7, 6, 5, 4, 3, 2, 1, or +3, +2, +1, 0, -1, -2, -3.
Many researchers find it desirable to assume that the semantic differential provides
interval data. This assumption, although widely accepted, has its critics, who argue that
the data have only ordinal properties because the weights are arbitrary. Depending on
whether the data are assumed to be interval or ordinal, the arithmetic mean or the
median is utilized to plot the profile of one concept, product, unit, etc., compared with
another concept, product, or units.
The semantic differential technique was originally developed by Charles Osgood and
others as a method for measuring the meaning of objects or the "semantic space" of
interpersonal experience." Business researchers have found the semantic differential
versatile and have modified it for business applications.
Numerical Scales
Numerical scales have numbers, rather than "semantic space" or verbal descriptions as
response options to identify categories (response positions). If the scale items have five
response positions, the scale is called a 5-point numerical scale; with seven response
positions, it is called a 7-point numerical scale, and so on.
Consider the following numerical scale:
Now that you've had your automobile for about one year, please tell us how satisfied
you are with your Ford Ikon:,
Extremely Satisfied
7
Extremely Dissatisfied
6
This numerical scale utilizes bipolar adjectives in the same manner as the semantic
differential.
Constant-Sum Scale
If a Parcel Service company wishes to determine the importance of the attributes of
accurate invoicing, delivery as promised, and price to organizations that use its service
in business-to-business marketing. Respondents might be asked to divide a constant
sum to indicate the relative importance of the attributes. For example:
Divide 100 points among the following characteristics of a delivery service according
to how important each characteristic is to you when selecting a delivery company.
Accurate invoicing___
Delivery as promised___
Lower price___
The constant-sum-scale works best with respondents with high educational levels. If
respondents follow instructions correctly, the results approximate interval measures. As
in the paired-comparison method, as the number of stimuli increases this technique
becomes more complex.
Stapel Scale
The Stapel scale was originally developed in the 1950s to measure the direction and
intensity of an attitude simultaneously. Modern versions of the scale use a single
adjective as a substitute for the semantic differential when it is difficult to create pairs of
bipolar adjectives. The modified Stapel scale places a single adjective in the center of an
even number of numerical values (for example, ranging from +3 to -3). It measures how
close to or how distant from the adjective a given stimulus is perceived to be.
The advantages and disadvantages of the Stapel scale are very similar to those of the
semantic differential. However, the Stapel scale is markedly easier to administer,
especially over the telephone. Because the Stapel scale does not requires bipolar
adjectives, as does the semantic differential, the Stapel scale is easier to construct.
Research comparing the semantic differential with the Stapel scale indicates that results
from the two techniques are largely the same.
Graphic Rating Scale
A graphic rating scale presents respondents with graphic continuum. The respondents
are allowed to choose any point on the continuum to indicate their attitudes. Typically, a
respondent's score is determined by measuring the length (in millimeters) from one end
of the graphic continuum to the point marked by the respondent. Many researchers
believe scoring in this manner strengthens the assumption that graphic rating scales of
this type are interval scales. Alternatively, the researcher may divide the line into
predetermined scoring categories (lengths) and record respondent's marks accordingly.
In other words, the graphic rating scale has the advantage of allowing the researchers to
choose any interval they wish for purposes of scoring. The disadvantage of the graphic
rating scale is that there are no standard answers.
Thurstone Equal-Appearing Interval Scale
In 1927, Louis Thurstone, an early pioneer in attitude research, developed the concept
that attitudes vary along continua and should be measured accordingly. Construction of
a Thurstone scale is a rather complex process that requires two stages. The first stage is
a ranking operation, performed by judges, who assigns scale values to attitudinal
statements. The second stage consists of asking subjects to respond to the attitudinal
statements.
The Thurstone method is time-consuming and costly. From a historical perspective it is
valuable, but its current popularity is low, because it is rarely utilized in most applied
business research.
Extremely likely
Very likely
Somewhat likely
Likely, about 50-50 chance
Somewhat unlikely
Very unlikely
Extremely unlikely
Behavioral Differential
A general instrument, the behavioral differential, has been developed to measure the
behavioral intentions of subjects toward an object or category of objects. As in the
semantic differential, a description of the object to be judged is placed on the top of a
sheet, and the subjects indicate their behavioral intentions toward this object on a series
of scales. For example, one item might be:
A 25-year-old female commodity broker
Would: ______:_____
Sorting tasks requires that respondents indicate their attitudes or beliefs by arranging
items.
SUMMARY
This chapter describes the technique for measuring attitude. This paper outlined the
importance of the attitude.
KEY TERMS
Attitude
Affective component
Cognitive component
Behavioral component
Ranking
Rating
Category scale
Likert scale
Semantic differential scale
Numerical scale
Constant sum scale
Stapel scale
Graphic rating scale
Paired comparison
QUESTIONS
1. What is an attitude?
2. Distinguish between rating and ranking. Which is a better attitude measurement?
Why?
3. Describe the different methods of scale construction, pointing out the merits and
demerits of each.
4. What advantages do numerical scales have over semantic differential scales?
OBJECTIVES
STRUCTURE
INTRODUCTION
Business researchers edit and code data to provide input that results in tabulated
information that will answer research question. With this input, the results can be
produced statistically and logically. Aspects of Statistics are important if the quantitative
data are to serve their purpose. If Statistics, as a subject, is inadequate and consists of
poor methodology, we would not know the right procedure to extract from the data the
information they contain. On the other hand, if our figures are defective in the sense
that they are inadequate or inaccurate, we would not reach the right conclusions even
though our subject is well developed. With this brief introduction, let us first see how
Statistics has been defined.
Major characteristics of statistics:
1. Statistics are aggregates of facts. This means that a single figure is not Statistics. For
example, national income of a country for a single year is not Statistics but the same for
two or more years is.
2. Statistics are affected by a number of factors. For example, sale of a product depends
on a number of factors such as its price, quality, competition, the income of the
consumers, and so on.
3. Statistics must be reasonably accurate. Wrong figures, if analyzed, will lead to
erroneous conclusions. Hence, it is necessary that conclusions must be based on
accurate figures.
4. Statistics must be collected in a systematic manner. If data are collected in a
haphazard manner, they will not be reliable and will lead to misleading conclusions.
5. Finally, Statistics should be placed in relation to each other. If one collects data
unrelated to each other, then such data will be confusing and will not lead to any logical
conclusions. Data should be comparable over tin e and over space.
Subdivisions in Statistics
The statisticians commonly classify this subject into two broad categories: descriptive
statistics and inferential statistics.
1. Descriptive Statistics
As the name suggests descriptive statistics includes any treatment designed to describe
or summaries the given data, bringing out their important features. These statistics do
not go beyond this. This means that no attempt is made to infer anything that pertains
to more than the data themselves. Thus, if someone compiles the necessary data and
reports that during the financial year 2000-2001, there were 1500 public limited
companies in India of which 1215 earned profits and the remaining 285 companies
sustained losses, his study belongs to the domain of descriptive Statistics. He may
further calculate the average profit earned per company as also average loss sustained
per company. This set of calculations too is a part of descriptive statistics.
Methods used in descriptive statistics may be called as descriptive methods. Under
descriptive methods, we learn frequency distribution, measures of central tendency, that
is, averages, measures of dispersion and skewness.
2. Inferential Statistics
Although descriptive Statistics is an important branch of Statistics and it continues to be
so, its recent growth indicates a shift in emphasis towards the methods of Statistical
inference. A few examples may be given here. The methods of Statistical inference are
required to predict the demand for a product such as tea or coffee for a company for a
specified year or years. Inferential Statistics are also necessary while comparing the
effectiveness of a given medicine in the treatment of any disease.
Again, while determining the nature and extent of relationship between two or more
variables like the number of hours studied by students and their performance in their
examinations, one has to take recourse to inferential Statistics.
Each of these examples is subject to uncertainty on account of partial, incomplete, or
indirect information. In such cases, the Statistician has to judge the merits of all possible
alternatives in order to make the most realistic prediction or to suggest the most
effective medicine or to establish a dependable relationship and the reasons for the
same. In this text, we shall first discuss various aspects of descriptive Statistics. This will
be followed by the discussion on different topics in inferential Statistics. The latter will
understandably be far more comprehensive than the former.
CENTRAL TENDENCY OF DATA
In many frequency distributions, the tabulated values show small frequencies at the
beginning and at the end and very high frequency at the middle of the distribution. This
indicates that the typical values of the variable lie near the central part of the
distribution and other values cluster around these central values. This behavior of the
data about the concentration of the values in the central part of the distribution is called
central tendency of the data. We shall measure this central tendency with the help of
mathematical quantities. A central value which 'enables' to comprehend in a single
effort the significance of the whole is known as Statistical Average or simply Average. In
fact, an average of a statistical series is the value of the variable which is representative
of the entire distribution and, therefore, gives a measure of central tendency.
Measures of Central Tendency
There are three common measures of central tendency
I. Mean
II. Median
III. Mode
The most common and useful measure of central tendency is, however the Mean. In the
following articles the method of calculation of various measures of central tendency will
be discussed. In all such discussion we need a very useful notation known as
Summation.
Choice of a Suitable Average
The different statistical average has different characteristics. There is no all-purpose
average. The choice of a particular average is usually determined by the purpose of
investigation. Within the framework of descriptive statistics, the main requirement is to
know what each average means and then select the one that fulfils the purpose at hand.
The nature of distribution also determines the type of average to be used.
Generally the following points should be kept in mind while making a choice of average
for use
1. Object
The average should be chosen according to the object of enquiry. If all the values in a
series are to be given equal importance, then arithmetic mean will be a suitable choice.
To determine the most stylish or most frequently occurring item mode should be found
out. If the object is to determine an average that would indicate its position or ranking in
relation to all the values, naturally, median should be the choice. If small items are to be
given greater importance than the big items, geometric mean is the best mean.
2. Representative
The average chosen should be such that it represents the basic characteristics of the
distribution.
3. Nature of form of data
If the frequently distribution is symmetrical or nearly symmetrical X, M or Mo may be
used almost interchangeably. If there are open-end class intervals, mean cannot be
calculated definitely. In a closed frequently distribution of unequal class intervals, it is
impossible to determine mode accurately. If there are a few values, it may not be
possible to determine mode. Mean will not give a representative picture, if there are few
extremely large or small values at either end of the array, and yet the great majority of
the values concentrate around a narrow band. In a variable of non-continuous type,
median or mode may give a value that actually exists in the data.
Davis' Test: Arithmetic mean is considered as an appropriate average for use for data
which has a symmetrical distribution or even if it has a moderate degree of asymmetry.
Prof. George Davis has devised a test which is:
If this coefficient works out to be more than +0.20, the distribution is symmetrical
enough to use arithmetic mean.
4. Characteristics of Average
While choosing a suitable average for a purpose, the merits and demerits of various
averages should always be considered and that average which fits into the purpose most
should be preferred over others. The following points should be given due consideration
in the process of selection of an average.
(i) In certain commonly encountered applications, the mean is subject to less sampling
variability than the median or mode.
(ii) Given only the original observations, the median is sometimes easiest to calculate.
Sometimes when there is no strong advantage for the mean, this advantage is enough to
indicate the use of the median.
(iii) Once a frequency distribution has been formed, the mode and the median are mode
quickly calculated than the mean. Moreover, when some classes are open-ended the
mean cannot be calculated from the frequency distribution.
(iv) The median is not a good measure when there are very few possible values for the
observations as with number of children or size of family.
(v) The mode and the median are relatively little affected by 'extreme' observations.
(vi) Calculations of geometric mean and harmonic mean is difficult as it involves the
knowledge of logarithms and reciprocals.
Hence "the justification of employing them (averages) must be determined by an appeal
to all the facts and in the light of the peculiar characteristics of the different types".
Uses of Different Averages
Different averages, due to their inherent characteristics are appropriate in different
circumstances. Thus, their use may be guided by the purpose at hand or circumstances
in which one is. Here a brief discussion is being made of uses of different statistical
averages:
1. Arithmetic Average
The arithmetic average used in the study of a social, economic or commercial problem
like production, income, price, imports, exports, etc. The central tendency if these
phenomena can best be studied by taking out an arithmetic average. Whenever we talk
of an 'average income' or 'average production' or 'average price' we always mean
arithmetic average of all these things. Whenever there is no indication about the type of
the average to be used, arithmetic average is computed.
2. Weighted Arithmetic Average
When it is desirable to give relative importance to the different items of a series,
weighted arithmetic average is computed. If it is desired to compute per capita
consumption of a family, due weights should be assigned to children, males and females.
This average is also useful in constructing numbers. The weighted average should be
used in the following cases:
a) If it is desired to have an average of whole group, which is divided into a
number of sub-classes, widely divergent from each other?
b) When items falling in various sub-classes change in such a way that the
proportion which the items bear among themselves also undergoes a change.
c) When combined average has to be computed.
d) When it is desired to calculate to find an average of ratios, percentages or rates.
3. Median
Median is especially applicable to cases which are not capable of precise quantitative
studies such as intelligence, honesty, etc. It is less applicable in economic or business
statistics, because there is lack of stability in such data.
4. Mode
The utility of mode is being appreciated more and more day by day. In r the sciences of
Biology and Meteorology it has been found to be of great value. In commerce and
industry it is gaining very great importance. Whenever as shop-keeper wants to stock
the goods he sells, he always looks to the modal size of those goods. Model size of shoes,
is of great importance to the businessman dealing in ready-made garments or shoes.
Many problems of production are related with mode. Many business establishments
these days are engaging their attention in keeping statistics of their sale ascertain the
particulars of the modal articles sold.
5. Geometric Mean
Geometric mean can advantageously be used in the construction of index numbers. It
makes the index numbers reversible and gives equal weight to equal ratio of changes.
This average is also useful in measuring the growth of population, because increases in
geometric progression. When there is wide dispersion in a serious, geometric mean is a
useful average.
6. Harmonic Mean
This average is useful in the cases where time, rate and prices are involved. When it is
desired to give the largest weight to the smallest item, this average is used.
Summation Notation ()
The symbol (read: sigma) means summation.
If x1, x2, x3, , xn be the n values of a variable x. Then their sum x1 +x2+ x3+ + xn is
shortly written as
Definition: x1, x2, x3, , xn be n values of a variable x, and if f1, f2, , fn be their respective
weights (or frequencies), then the weighted arithmetic mean is defined by
Measures of Skewness
The degree of skewness is measured by its coefficient. The common measures of
skewness are:
1. Pearson's first measure
Skewness = (Mean Mode) / Standard Deviation
where Q1, Q2, Q3 are the first, second and third quartiles respectively.
4. Moment Measure
Skewness = m3 / 3 = m3 / m23/2
where m2 and m3 are the second and third central moments and is the S.D.
All the four measure of Skewness defined above are independent of the units of
measurement.
Example:
Calculate the Pearson's measure of Skewness on the basis of Mean, Mode, and Standard
Deviation.
Solution:
According to Pearson's first measure, Skewness = (Mean Mode) / Standard Deviation
Here mid-values of the class-intervals are given. Assuming a continuous series, we
construct the following table:
greater than the median, and the median is greater than the mode (M > Me > Mo), and
the difference between upper quartile and median is greater than the difference between
median and lower quartile (Q3-Me > Me-Q1). In a negatively skewed distribution, more
than half of the area under the distribution curve is to the left side of the mode. In such a
distribution the elongated tail us to the left and mean is less than the median and
median is less than the mode (M < Me < Mo), and the difference between upper quartile
and median is less than the difference between median and lower quartile (Q3-M < MQ1).The following figures show these facts. The following table will also show these facts
of Position of Average on Various Distributions.
Test of Skewness
In order to find out whether a distribution is symmetrical or skewed, the following facts
should be noticed:
1. Relationship between Averages
If in a distribution mean, median and mode are not identical, then it is a skewed
distribution. The greater is the difference between mean and mode more will be the
skewness in the distribution.
2. Total of Deviations
If the sum of positive deviations from median or mode is equal to the sum of negative
deviation, then is no skewness in the distribution. The extent of difference between the
sums of positive and negative deviations from median or mode will determine the extent
of skewness in the data.
3. The distance of partition values from median
In a symmetrical distribution Q1 and Q2, D1 and D9 and P10 and P90 are equidistant from
median. In an asymmetrical distribution it is not so.
4. The frequencies on either side of the mode
In an asymmetrical distribution, frequencies on either side of the mode are not equal.
5. The curve
When the data are plotted on a graph paper, the curve will not be bell-shaped, or when
cut along a vertical line through the centre, the two halves will not be identical.
Conversely stated, in the absence of skewness in the distribution:
(i) Values of mean, median and mode will coincide.
(ii) Sum of the positive deviations from the median or mode will be equal to the sum of
negative deviations.
(iii) The two quartiles, deciles one and nine, and percentile ten and ninety will be
equidistant from the median.
(iv) Frequencies on the either side of the mode will be equal.
(v) Data when plotted on a graph paper will take a bell-shaped form.
Measures of Skewness
To find out the direction and the extent of symmetry in a series statistical measures of
skewness are calculated, these measures can be absolute or relative. Absolute measures
of skewness tell us the extent of asymmetry and whether it is positive or negative. The
absolute skewness can be known by taking the deference between mean and mode.
Symbolically,
Absolute SK = X - Mo
If the value of mean is greater than the mode (M > Mo) skewness will be positive. In
case the value of mean is less than the mode (M < Mo) skewness will be negative. The
greater is the amount of skewness, the more the mean and mode differ because of the
influence of extreme items. The reason why the difference between mean and mode is
taken for the measure of skewness is that in a symmetrical distribution, both the values
along with median coincide, but in an asymmetrical distribution, there will be a
difference between the mean and mode.
Thus the difference between the mean and the mode, whether positive or negative,
indicates that the distribution is asymmetrical. However such absolute measure of
skewness is unsatisfactory, because:
(1) It cannot be used for comparison of skewness in tow distributions if they are in
different units, because the difference between the mean and the mode will be in terms
of the units of distribution.
(2) The difference between the mean and mode may be more in one series and less in
another, yet the frequency curves of the two distributions may be similarly skewed. For
comparison, the absolute measures of skewness are changed to relative measures, which
are called Coefficient of Skewness.
There are four measures of relative skewness. They are:
1. The Karl Pearson's Coefficient of Skewness
2. The Bowley's Coefficient of Skewness.
3. The Kelly's Coefficient of Skewness
4. Measure of skewness based onmoments. Measures of Skewness
1. The Karl Pearson's Coefficient of Skewness
Karl Pearson has given a formula, for relative measure of Skewness. It is known as Karl
Pearson's Coefficient of Skewness or Pearsonian Coefficient of Skewness. The formula is
based on the difference between the mean and mode divided by the standard deviation.
The coefficient is represented by J
The Pearsonian coefficient of skewness has the interesting characteristic that it will be
positive when the mean is larger than the mode or median, and it will be negative when
the arithmetic mean is smaller than the mode or median. In a symmetrical distribution,
the value of Pearsonian coefficient of skewness will be zero.
There is no theoretical limit to this measure, however, in practical the value given by this
formula is rarely very high and usually lies between . The direction of the skewness is
given by the algebraic sign of the measure; if it is plus then the skewness is positive, if it
is minus, the skewness is negative. The degree of skewness is obtained by the numerical
figure such as 0.9, 0.4, etc.
Thus this formula gives both the direction as well as the degree of skewness. There is
another relative measure of skewness also based on the position of averages. In this, the
difference between two averages is divided by the mean deviation. The formula is:
These formulas are not very much used in practice, because of demerits of mean
deviation.
DISPERSION
Measures of Dispersion
An average nay give a good idea of the type of data, but it alone can't reveal all the
characteristics of data. It cannot tell us in what manner all the values of the variable are
scattered / dispersed about the average.
Meaning of Dispersion
The Variation or Scattering or Deviation of the different values of a variable from their
average is known as Dispersion. Dispersion indicates the extent to which the values vary
among themselves. Prof. W.I. King defines the term, 'Dispersion' as it is used to indicate
the facts that within a given group, the items differ from another in size or in other
words, there is lack of uniformity in their sizes. The extent of variability in a given set of
data is measured by comparing the individual values of the variable with the average all
the values and then calculating the average of all the individual differences.
Objectives of Measuring Variations
1. To serve as a basis for control of the variability itself.
2. To gauge the reliability of an average
3. To serve as a basis for control of the variability itself
Types of Measures of Dispersion
There are two types of measures of dispersion. The first, which may be referred to as
Distance Measures, describes the spread of data in terms of distance between the values
of selected observations. The second are those which are in terms of an average
deviation from some measure of central tendency.
Absolute and Relative Measures of Dispersion
Measures of absolute dispersion are in the same units as the data whose scatter they
measure. For example, the dispersion of salaries about an average is measured in rupees
and the variation of time requires for workers to do a job is measured in minutes or
hours. Measures of absolute dispersion cannot be used to compare the scatter in one
distribution with that in another distribution when the averages of the distributions
differ in size or the units of measure differ in kind. Measures of relative dispersion show
some measure of scatter as a percentage or coefficient of the absolute measure of
dispersion. They are generally used to compare the scatter in one distribution with the
scatter in other. Relative measure of dispersion is called coefficient of dispersion.
Methods of Measuring Dispersion
There are two meanings of dispersion, as explained above. On the basis of these two
meanings, there are two mathematical methods of finding dispersion, i.e. methods of
limits and methods of moment. Dispersion can also be studied graphically. Thus, the
following are the methods of measuring dispersion:
I. Numerical Methods
1. Methods of Limits
i. The Range
ii. The Inter-Quartile Range
Example:
Find the quartile deviation of the following frequency distribution.
Daily wages:
No of workers:
10-15
15-20
20-25
25-30
30-35
12
18
10
Solution:
Note: The expression |d| is read as mod. d and gives only numerical or absolute value of
d without regard to sign. Thus,
|-3| = 3, |-4| = 4, |0.56| = 0.56
The reason for taking only the absolute and not the algebraic values of the deviation is
that the algebraic sum of the deviations of the value from their mean is zero.
Example:
Find the Mean Deviation about the Arithmetic Mean of the numbers 31, 35, 29, 63, 55,
72, 37.
Solution:
Arithmetic Mean = (31 + 35 + 29 + 63 + 55 + 72 + 37) / 7 = 322/7 = 46
Calculation of absolute Deviations
The required Mean Deviation about the Mean = |d| / n = 104 / 7 = 14.86
Advantages
Mean deviation is based on all the values of the variable and sometimes gives fairly good
result as a measure of dispersion. However the practice of neglecting signs and taking
absolute deviations for the calculation of the mean Deviation seems rather unjustified
and this makes algebraic treatment difficult.
STANDARD DEVIATION
It is the most important absolute measures of Dispersion. Standard Deviation of set
values of a variable is defined as the positive square root of arithmetic mean of the
squares of all deviations of the values from their arithmetic mean. In short it is the
square root of the mean of the squares of deviations from mean.
If x1, x2, ...,xn be a series of values of a variable and x bar their AM, then S.D () is
defined by
Standard Deviation is the most important and widely used among the measures of
dispersion and it possesses almost all the requisites of a good measure of dispersion. It
is rigidly defined and based on all the values of the variable.
It is suitable for algebraic treatment. SD is less affected by sampling fluctuations than
any other absolute measure of dispersion.
SD is difficult to understand. The process of squaring the deviations from their AM and
then taking the square-root of the AM of these squared deviations is a complicated
affair.
The calculation of SD can be made easier by changing the origin and the scale
conveniently.
Relative Measures of Dispersion
Absolute measures expressed as a percentage of a measure of a control tendency gives
relative measures of dispersion. Relative measures are independent of the units of
measurement and hence they are used for the comparison of dispersion of two or more
distributions given in different units.
Co-efficient of Variation
Co-efficient of variation is the first important relative measure of dispersion and is
defined by the following formula:
Co-efficient of variation = Standard Deviation / Mean x 100
Co-efficient of variation is thus the ratio of the Standard deviation to the mean,
expressed as a percentage. In the words of Karl Pearson, Co-efficient of Variation is the
percentage variation in the mean.
Example:
Find the Mean Deviation about the Median in respect of the following numbers: 46, 79,
26, 85, 39, 65, 99, 29, 56, and 72. Find also the Co efficient of Mean Deviation.
Solution:
By arranging the given numbers in ascending order of magnitude, we obtain 26, 29, 39,
46, 56, 65, 72, 79, 85, and 99.
Median = [(n+1)/2]th value = [11/2]th value = 5.5th value = (5th value + 6th value) / 2 =
(56+65)/2 = 60.5
Absolute deviation of the values from the Median 60.5 is respectively
34.5, 31.5, 21.5, 14.5, 4.5, 4.5, 11.5, 18.5, 24.5, and 38.5
Therefore, Mean Deviation (MD) about the Median
=(34.5+31.5+21.5+14.5+4.5+4.5+11.5+18.5+24.5+38.5)/2 = 20.4
Co-efficient of MD = MD / Median x 100 = 20.4 / 60.5 x 100 = 33.72%
KURTOSIS
Kurtosis in Greek means bulginess. The degree of kurtosis of a distribution is measured
relative to the peakedness of a normal curve. The measure of kurtosis indicates whether
the curve of the frequency distribution is flat or peaked.
Kurtosis is the peakedness of a frequency curve. In two or more distributions having
same average, dispersion and skewdness, one may have higher concentration of values
near the mode, and its frequency curve will show sharper peak than the others. This
characteristic of frequency distribution is known as Kurtosis.
Kurtosis is measured by the coefficient 2, which is defined by the formula
2 = m4 / m22 = m4 / 4 or by 2 = 2 3, where m4 and m2 are the 2nd and 4th control
moments, and = SD.
A distribution is said to be Platy-kurtic, Meso-kurtic, and Lepto-kurtic corresponding to
2 < 3, 2 = 3, or 2 > 3.
Accordingly,
Its Platy-kurtic, if m4 / 4 < 3
Its Meso-kurtic, if m4 / 4 = 3
Its Lepto-kurtic, if m4 / 4 > 3
Karl person in 1905 introduced the terms MESOKURTIC, LEPTOKURTIC, and
PLATYKURTIC. A peaked curve is called "Leptokurtic" and a flat topped curve is termed
"Platykurtic". These are evaluated by comparison with intermediate peaked curve. These
three curves differ widely in regard to convexity.
Example:
Calculate the measures of the following distribution:
Solution:
V1 = fd / N = 0/100 = 0
V2 = fd2 / N = 446/100 = 4.46
V3 = fd3 / N = -36/100 = -0.36
V4 = fd4 / N = 4574/100 = 45.74
4 = V4 4V1V3 + 6V12V2 3V14
= 45.74 4(0)(-0.36) + 6(0)2(4.46) 3(0)4
4 or 2 = V4/ V22 = 45.74 / 4.462 = 2.3
The value of 2 is less than 3, hence the curve is Platykurtic.
SUMMARY
This chapter helps us to know the nature of the statistical study. This chapter recognizes
the importance of statistics and also its limitations. The differences between descriptive
statistics and inferential statistics are dealt in detail.
KEY WORDS
Descriptive Statistics
Inferential Statistics
Mean
Standard Deviation
Median
Mean Deviation
Mode
Arithmetic Mean
Geometric Mean
Averages
Dispersion
Skewness
Mean deviation
Kurtosis
REVIEW QUESTIONS
1. Explain the special features of measures of central tendency.
2. How will you choose an 'average'?
3. What is dispersion? State its objectives. Explain the various types measures of
dispersion.
4. Explain the various methods of measuring dispersion.
5. Differentiate standard deviation from mean deviation.
6. Define 'skewdness'. How will you measure it?
7. Explain the application of averages in research.
OBJECTIVES
To study simple, partial and multiple correlation and their application in research
STRUCTURE
Measure of relationship
Correlation
Properties of correlation co-efficient
Methods of studying correlations
Application of correlation
MEASURES OF RELATIONSHIP
The following statistical tools measure the relationship between the variables analyzed
in social science research:
(i) Correlation
Simple correlation
Partial correlation
Multiple correlation
(ii) Regression
Simple regression
Multiple regressions
Frequently correlation analysis is also used along with the regression analysis to
measure how well the regression line explains the variations of the dependent variable.
The correlation coefficient is the statistical tool that is used to measure the mutual
relationship between the two variables".
Coefficient of correlation is denoted by 'r'.
The sign of 'r' shows the direction of the relationship between the two variables X and Y.
Positive correlation reveals that there is a positive correlation between the two variables.
Negative correlation reveals negative relationship. Levin states that if an inverse
relationship exists that is, if Y decreases as X increases then r will fall between 0
and -1. Likewise, if there is a direct relationship (if Y increases as X increases), then r
will be a value within the range of 0 to 1.
Aggarwal Y.P. has highlighted in his book 'Statistical Methods', the properties of
correlation and factors influencing the size of the correlation coefficient. The details are
given below:
PROPERTIES OF THE CORRELATION COEFFICIENT
The range of correlation coefficient is from -1 through 0 to +1. The values of r = -1 and r
= +1 reveal a case of perfect relationship, though the direction of relationship is negative
in the first case, and positive in the second case.
The correlation coefficient can be interpreted in terms of r2. It is known as 'coefficient of
determination'. It may be considered as the variance interpretation of r2.
Example:
r = 0.5
r2 = 0.5 x 0.5 = 0.25. In terms of percentage, 0.25x100% = 25%
It refers that 25 percent of the variance in Y scores has been accounted for by the
variance in X.
The correlation coefficient does not change if every score in either or both distribution is
increased or multiplied by a constant.
Causality cannot be inferred solely as the basis of a correlation between two variables. It
can be inferred only after conducting controlled experiments.
The direction of the relation is indicated by the sign (+ or -) of the correlation.
The degree of relationship is indicated by the numerical value of the correlation. A value
near 1 indicates a nearly perfect relation, and a value near 0 indicates no relationship.
variables are studied, but the effect on one variable is kept constant and relationship
between other two variables is studied
c) Linear and Non-Linear Correlations
It depends upon the constancy of the ratio of change between the variables. In linear
correlation the percentage change in one variable will be equal to the percentage change
in another variable. It is not so in non-linear correlation.
METHODS OF STUDYING CORRELATION
a) Scatter Diagram Method
b) Graphic Method
c) Karl Pearson's Coefficient of Correlation
d) Concurrent Deviation Method
e) Method of Least Squares
Karl Pearson's Coefficient of Correlation
Procedure
i. Compute mean of the X series data
ii. Compute mean of the Y series data
iii. Compute deviations of X series from the mean of X. It is denoted as x.
iv. Square the deviations. It is denoted as x2.
v. Compute deviations of Y series from the mean of Y. It is denoted as y.
vi. Square the deviations. It is denoted as y2.
vii. Multiply deviation (X series, Y series) and compute total. It denoted as xy.
The above values can be applied in the formula and correlation can be computed.
Karl Pearson's Coefficient of Correlation (r)
where,
dx = sum of deviations of X series from the assumed mean
dy = sum of deviations of Y series from the assumed mean
dxdy = total of deviations (X and Y series)
dx2 = deviations of X series from assumed mean are squared
dy2 = deviations of Y series from assumed mean are squared
N = Number of items
The above values can be applied in the above formula and correlation can be computed.
Correlation for the grouped data can be computed with the help of the following
formula:
In the above formula, deviations are multiplied by the frequencies. Other steps are the
same.
CALCULATION OF CORRELATION
Raw Score Method
r = 0.7
Solution
Karl Pearsons Correlation Coefficients:
Year
1985
Index of
Production (X)
100
x = (XXmean)
-4
x2
16
y2
xy
1986
1987
1988
1989
1990
1991
1992
102
104
107
105
112
103
99
X
= 832
-2
0
+3
+1
+8
-1
-5
x
=0
4
0
9
1
64
1
25
x2
= 120
12
13
11
12
12
19
26
Y
= 120
-3
-2
-4
-3
-3
+4
+11
y
=0
= X / N = 832/8 = 104
= Y / N = 120/8 = 15
Solution
9
+6
4
0
16
-12
9
-3
9
-24
16
-4
121
-55
2
y
xy
= 184 = -92
RANK CORRELATION
It is a method of ascertaining co variability or the lack of it between the two variables.
Rank correlation method is developed by the British Psychologist Charles Edward
Spearmen in 1904. Gupta S.P has stated that "the rank correlation method is used when
quantitative measures for certain factors cannot be fixed, but individual in the group can
be arranged in order thereby obtaining for each individual a number of indicating
his/her rank in the group".
The formula for Rank Correlation
Rank Correlation,
Testing of Correlation
't' test is used to test correlation coefficient. Height and weight of a random sample of six
adults is given.
It is reasonable to assume that these variables are normally distributed, so the Karl
Pearson Correlation coefficient is the appropriate measure of the degree of association
between height and weight.
r = 0.875
Hypothesis test for Pearson's population correlation coefficient
H0: p = 0 - this implies no correlation between the variables in the population
H1: p > 0 - this implies that there is positive correlation in the population (increasing
height is associated with increasing weight)
5% significance level
variable is partially out from the correlation between other two variables. This statistical
technique is known as partial correlation.
Correlation between variables x and y is denoted as rxy
Partial correlation is denoted by the symbol r123. This is correlation between variables 1
and 2, keeping 3rd variable constant.
where,
r123 = partial correlation between variables 1 and 2
r12 = correlation between variables 1 and 2
r13 = correlation between variables 1 and 3
r23 = correlation between variables 2 and 3
Multiple Correlation
Three or more variables are involved in multiple correlation. The dependent variable is
denoted by X1 and other variables are denoted by X2, X3 etc. Gupta S. P. has expressed
that "the coefficient of multiple linear correlation is represented by R1 and it is common
to add subscripts designating the variables involved. Thus R1.234 would represent the
coefficient of multiple linear correlation between X1 on the one hand, X2, X3 and X4 on
the other. The subscript of the dependent variable is always to the left of the point".
The coefficient of multiple correlation for r12, r13 and r23 can be expressed as follows:
Coefficient of multiple correlations for R1.23 is the same as R1.32. A coefficient of multiple
correlation lies between 0 and 1. If the coefficient of multiple correlations is 1, it shows
that the correlation is perfect. If it is 0, it shows that there is no linear relationship
between the variables. The coefficients of multiple correlation are always positive in sign
and range from +1 to 0.
Coefficient of multiple determinations can be obtained by squaring R1.23.
Multiple correlation analysis measures the relationship between the given variables. In
this analysis the degree of association between one variable considered as the dependent
variable and a group of other variables considered as the independent variables.
SUMMARY
This chapter outlined the significance in measuring the relationship. This chapter
discuss the factors that affecting correlation. The different applications of correlation
have been dealt in detail.
KEY WORDS
Measures of Relationship
Correlation
Simple correlation
Partial correlation
Multiple correlation
Regression
Simple regression
Multiple regressions
Association of Attributes
Scatter Diagram Method
Graphic Method
REVIEW QUESTIONS
1. What are the different measures and their significance in measuring Relationship?
2. Discuss the factors affecting Correlation.
3. What are the applications of Correlation?
4. Discuss in detail on different types of Correlation.
REFERENCE BOOKS
1. Robert Ferber, Marketing research, New York: McGraw Hill Inc., 1976.
2. Chaturvedhi, J.C., Mathematical Statistics, Agra: Nok Jhonk Karyalaya, 1953.
3. Emony, C. William, Business Research Methods, Illinois, Irwin, Homewood. 1976.
STRUCTURE
Table
Relations frequency table
Cross tabulation and stub-and-banner tables
Guideline for cross tabulation
INTRODUCTION
To get meaningful information from the data it is arranged in the tabular form.
Frequency tables, histograms are simple form of tables.
Frequency Tables
Frequency table or frequency distribution is a better way to arrange data. It helps in
compressing data. Though some information is lost, compressed data show a pattern
clearly. For constructing a frequency table, the data are divided into groups of similar
values (class) and then record the number of observations that fall in each group.
Table 1: Frequency table on age-wise classification of respondents
The data of collection days are presented in the following table as a frequency table. The
number of classed can be increased by reducing the size of the class. The choice of class
intervals is mostly guided by practical consideration rather than by rules. Class intervals
are made in such a way that measurements are uniformly distributed over the class and
the interval is not very large. Otherwise, the mid value will either overestimate or
underestimate the measurement.
Relative frequency tables
Frequency is total number of data points that fall within that class. Frequency of each
value can also be expressed as a fraction or percentage of the total number of
observations. Frequencies expressed in percentage terms are known as relative
frequencies. A relative frequency distribution is presented in the table below.
Table 2: Relative frequency table on occupation-wise classification of
respondents
It may be observed that the sum of all relative frequencies is 1.00 or 100% because
frequency of each class has been expressed as a percentage of the total data.
The table above shows the number, proportion, and cumulative proportion of
respondents who characterized their interest in watching football as either (1) Always
interested, (2) Usually interested, (3) Sometimes interested, or (4) Never interested.
Applications
In practically every research project, a first "look" at the data usually includes frequency
tables. For example, in survey research, frequency tables can show the number of males
and females who participated in the survey, the number of respondents from particular
ethnic and racial backgrounds, and so on. Responses on some labeled attitude
measurement scales (e.g., interest in watching football) can also be nicely summarized
via the frequency table. In medical research, one may tabulate the number of patients
displaying specific symptoms; in industrial research one may tabulate the frequency of
different causes leading to catastrophic failure of products during stress tests (e.g.,
which parts are actually responsible for the complete malfunction of television sets
under extreme temperatures?). Customarily, if a data set includes any categorical data,
then one of the first steps in the data analysis is to compute a frequency table for those
categorical variables.
Cross Tabulation and Stub-and-Banner Tables
Managers and researchers frequently are interested in gaining a better understanding of
the differences that exist between two or more subgroups. Whenever they try to identify
characteristics common to one subgroup but not common to other subgroups, (i.e. they
are trying to explain differences between the subgroups). Cross tables are used to
explain the difference between the subgroups.
Cross tabulation is a combination of two (or more) frequency tables arranged such that
each cell in the resulting table represents a unique combination of specific values of
cross tabulated variables. Thus, cross tabulation allows us to examine frequencies of
observations that belong to specific categories on more than one variable.
By examining these frequencies, we can identify relations between cross tabulated
variables. Only categorical variables or variables with a relatively small number of
different meaningful values should be cross tabulated. Note that in the cases where we
do want to include a continuous variable in a cross tabulation (e.g., income), we can first
recode it into a particular number of distinct ranges (e.g. low, medium, high).
Guidelines for Cross Tabulation
The most commonly used method of data analysis is cross tabulation. The following
guidelines will helpful to design proper cross tabulation,
1. The data should be in categorical form
Cross tabulation is applicable to data 1 which both the dependent and the independent
variables appear in categorical form. There are two types of categorical data.
One type (assume type A) consists of variables that can be measured only in classes or
categories. Like marital status, gender, occupation variables can be measured in
categories not quantifiable (i.e. no measurable number).
Another type (say type B) variables, which can be measured in numbers, such as age,
income. For this type the different categories are associated with quantifiable numbers
that show a progression from smaller values to larger values.
Cross tabulation is used on both types of categorical variables. However when
construction across tabulation is done using type B categorical variables, researchers
find it helpful to use several special steps to make such cross tabulations more effective
analysis tools.
1. If certain variable is believed to be influenced by some other variable, the former can
be considered to be a dependent variable and the later is called as independent
variable.
2. Cross tabulate an important dependent variable with one or more
'explaining' independent variables.
Researchers typically cross tabulate a dependent variable of importance to the objectives
of the research project (such as heavy user versus light user or positive attitude versus
negative attitude) with one or more independent variables that the researchers believe
can help explain the variation observed in the dependent variable. Any two variables can
be used in a cross tabulation so long as they both are in categorical form, and they both
appear to be logically related to one another as dependent and independent variables
consistent with the purpose and objectives of the research project.
3. Show percentage in a cross tabulation
In a cross tabulation researchers typically show the percentage as well as the actual
count s of the number of responses falling into the different cells of the table. The
percentages more effectively reveal the relative sizes of the actual counts associated with
the different cells and make it easier for researchers to visualize the patterns of
differences that exist in the data.
Constructing and Interpreting a Cross Tabulation
After drawing the cross table the interpretations has to be drawn from the table. It
should convey the meaning and findings from the table. In management research
interpretations has more value. From the interpretations and findings managers take
decisions.
2x2 Tables
The simplest form of cross tabulation is the 2 by 2 table where two variables are
"crossed," and each variable has only two distinct values. For example, suppose we
conduct a simple study in which males and females are asked to choose one of two
different brands of soda pop (brand A and brand B); the data file can be arranged like
this:
Each cell represents a unique combination of values of the two cross tabulated variables
(row variable Gender and column variable Soda), and the numbers in each cell tell us
how many observations fall into each combination of values. In general, this table shows
us that more females than males chose the soda pop brand A, and that more males than
females chose soda B. Thus, gender and preference for a particular brand of soda may be
related (later we will see how this relationship can be measured).
Marginal Frequencies
The values in the margins of the table are simply one-way (frequency) tables for all
values in the table. They are important in that they help us to evaluate the arrangement
of frequencies in individual columns or rows. For example, the frequencies of 40% and
60% of males and females (respectively) who chose soda A (see the first column of the
above table), would not indicate any relationship between Gender and Soda if the
marginal frequencies for Gender were also 40% and 60%; in that case they would simply
reflect the different proportions of males and females in the study. Thus, the differences
between the distributions of frequencies in individual rows (or columns) and in the
respective margins inform us about the relationship between the cross tabulated
variables.
Column, Row, and total Percentages. The example in the previous paragraph
demonstrates that in order to evaluate relationships between cross tabulated variables
we need to compare the proportions of marginal and individual column or row
frequencies. Such comparisons are easiest to perform when the frequencies are
presented as percentages.
Evaluating the Cross Table
Researchers find it useful to answer the following three questions when evaluating cross
tabulation that appears to explain differences in a dependent variable.
1. Does the cross tabulation show a valid or a spurious relationship?
2. How many independent variables should be used in the cross tabulation?
3. Are the differences seen in the cross tabulation statistically significant, or could they
have occurred by chance due to sampling variation?
Class
Frequency
Relations frequency
Cumulative frequency
Marginal frequency
Cross table
REVIEW QUESTIONS
1. Why do we use cross tables?
2. How do you evaluate the cross table?
3. Define the guidelines for constructing the cross table
OBJECTIVES
To learn the application of various statistical packages used for the management
research process
To understand the procedures for performing the tests using SPSS
STRUCTURE
Statistical packages
Statistical analysis using SPSS
t-test, F-test, chi-square test, Anova
Factor analysis
Statistical Packages
The following statistical software packages are widely used:
STATA
SPSS
SAS
STATA
Stata, created in 1985 by Statacorp, is a statistical program used by many businesses and
academic institutions around the world. Most of its users work in research, especially in
the fields of economics, sociology, political science, and epidemiology.
Stata's full range of capabilities includes:
Data management
Statistical analysis
Graphics
Simulations
Custom programming
SPSS
The computer program SPSS (originally, Statistical Package for the Social Sciences) was
released in its first version in 1968, and is among the most widely used programs for
statistical analysis in social science. It is used by market researchers, health researchers,
survey companies, government, education researchers' and others. In addition to
statistical analysis, data management (case selection, file reshaping, creating derived
data) and data documentation are features of the base software.
The many features of SPSS are accessible via pull-down menus (see image) or can be
programmed with a proprietary 4GL "command syntax language". Command syntax
programming has the benefits of reproducibility and handling complex data
manipulations and analyses
Solve business and research problems using SPSS for Windows, a statistical and data
management package for analysts and researchers.
SPSS for Windows provides you with a broad range of capabilities for the entire
analytical process. With SPSS, you can generate decision-making information quickly
using powerful statistics, understand and effectively present the results with highquality tabular and graphical output, and share the results with others using a variety of
reporting methods, including secure Web publishing. Results from the data analysis
enable you to make smarter decisions more quickly by uncovering key facts, patterns,
and trends. An optional server version delivers enterprise-strength scalability,
additional tools, security, and enhanced performance.
SPSS can be used for Windows in a variety of areas, including:
In particular, apply SPSS statistics software to gain greater insight into the actions,
attributes, and attitudes of peopleth - e customers, employees, students, or citizens.
Add more functionality as you need it
SPSS for Windows is a modular, tightly integrated, full-featured product line for the
analytical process planning, data collecting, data access, data management and
preparation, data analysis, reporting, and deployment. Using a combination of add-on
modules and stand-alone software that work seamlessly with SPSS Base enhances the
capabilities of this statistics software. The intuitive interface makes it, easy to use - yet it
gives you all of the data management, statistics, and reporting methods you need to do a
wide range of analysis.
Gain unlimited programming capabilities
Dramatically increase the power and capabilities of SPSS for Windows by using the
SPSS Programmability Extension. This feature enables analytic and application
developers to extend the SPSS command syntax language to create procedures and
applications - and perform even the most complex jobs within SPSS. The SPSS
Programmability Extension is included with SPSS Base, making this statistics software
an even more powerful solution.
Maximize market opportunities
The more competitive and challenging the business environment, the more you need
market research. Market research is the systematic and objective gathering, analysis,
and interpretation of information. It helps the organization identify problems and
opportunities and allows for better-informed, lower-risk decisions.
For decades, solutions from SPSS Inc. have added value for those involved in market
research. SPSS solutions support the efficient gathering of market research information
through many different methods, and make it easier to analyze and interpret this
information and provide it to decision makers.
We offer solutions both to companies that specialize in providing market research
services and to organizations that conduct their own market research. SPSS market
research solutions help you:
With this insight, you or the clients can confidently make decisions about developing
and marketing the products and enhancing the brand.
The heart of SPSS market research solution is dimensions product family. Through
Dimensions, the organization can centralize the creation and fielding of surveys in any
mode and in any language, as well as the analysis and reporting phases of the research.
Dimensions data can be directly accessed using SPSS for Windows, which enables the
analysts to use SPSS' advanced statistical and graphing capabilities to explore the survey
data. Add-on modules and integrated stand-alone products extend SPSS' analytical and
reporting capabilities. For example, analyze responses to open-ended survey questions
with SPSS Text Analysis for Surveys.
Maximize the value the organization receives from its Dimensions data by using an
enterprise feedback management (EFM) solution from SPSS. EFM provides you with a
continuous means of incorporating regular customer insight into the business
operations. Engage with current or prospective customers through targeted feedback
programs or by asking questions during naturally occurring events. Then use the
resulting insights to drive business improvement across the organization. SPSS' EFM
solution also enables you to integrate the survey data with transactional and operational
data, so you gain a more accurate, complete understanding of customer preferences,
motivations, and intentions.
Thanks to the integration among SPSS offerings, you can incorporate insights gained
through survey research in the predictive models created by the data mining tools. You
can then deploy predictive insight and recommendations to people and to automated
systems through any of the predictive analytics applications.
SAS
The SAS System, originally Statistical Analysis System, is an integrated system of
software products provided by SAS Institute that enables the programmer to perform:
In addition, the SAS System integrates with many SAS business solutions that enable
large scale software solutions for areas such as human resource management, financial
management, business intelligence, customer relationship management and more.
Statistical analyses using SPSS
Introduction
This section shows how to perform a number of statistical tests using SPSS. Each section
gives a brief description of the aim of the statistical test, when it is used, an example
showing the SPSS commands and SPSS (often abbreviated} output with a brief
interpretation of the output. In deciding which test is appropriate to use, it is important
to consider the type of variables that you have (i.e., whether your variables are
categorical, ordinal or interval and whether they are normally distributed).
Statistical methods using SPSS
One sample t-test
A one sample t-test allows us to test whether a sample mean (of a normally distributed
interval variable) significantly differs from a hypothesized value. For example, using the
data file, say we wish to test whether the average writing score (write) differs
significantly from 50. We can do this as shown below:
t-test
/testval = 50
/variable = write.
The mean of the variable write for this particular sample of students is 52.775, which is
statistically significantly different from the test value of 50. We would conclude that this
group of students has a significantly higher mean on the writing test than 50.
One sample median test
A one sample median test allows us to test whether a sample median differs significantly
from a hypothesized value. We will use the same variable, write, as we did in the one
sample t-test example above, but we do not need to assume that it is interval and
normally distributed (we only need to assume that write is an ordinal variable).
However, we are unaware of how to perform this test in SPSS.
Binomial test
A one sample binomial test allows us to test whether the proportion of successes on a
two-level categorical dependent variable significantly differs from a hypothesized value.
For example, using the, say we wish to test whether the proportion of females (female)
differs significantly from 50%, i.e., from 0.5. We can do this as shown below:
npar tests
/binomial (.5) = female.
a. Based on Z Approximation.
The results indicate that there is no statistically significant difference (p =0.229). In
other words, the proportion of females in this sample does not significantly differ from
the hypothesized value of 50%.
Chi-square goodness of fit
A chi-square goodness of fit test allows us to test whether the observed proportions for a
categorical variable differ from hypothesized proportions. For example, let's suppose
that we believe that the general population consists of 10% Hispanic, 10% Asian, 10%
African American and 70% White folks. We want to test whether the observed
proportions from our sample differ significantly from these hypothesized proportions.
npar test
/chisquare = race
/expected = 10 10 10 70.
a. 0 cells (0%) have expected count less than 5. The minimum expected cell frequency is
20.0
These results show that racial composition in our sample does not differ significantly
from the hypothesized values that we supplied (chi-square with three degrees of
freedom = 5.029, p = 0.170).
Two independent samples t-test
An independent samples t-test is used when you want to compare the means of a
normally distributed interval dependent variable for two independent groups. For
example, using the, say we wish to test whether the mean for write is the same for males
and females.
t-test groups = female(0 1)
/variables = write.
The results indicate that there is a statistically significant difference between the mean
writing score for males and females (t = -3.734, p = 0.000). In other words, females
have a statistically significantly higher mean score on writing (54.99) than males
(50.12).
Chi-square test
A chi-square test is used when you want to see if there is a relationship between two
categorical variables. In SPSS, the chi2 option is used with the tabulate command to
obtain the test statistic and its associated p-value. Let's see if there is a relationship
between the type of school attended (schtyp) and students gender (female).
Remember that the chi-square test assumes that the expected value for each cell is five
or higher. This assumption is easily met in the examples below. However, if this
assumption' is not met in your data, please see the section on Fisher's exact test below.
These results indicate that there is no statistically significant relationship between the
type of school attended and gender (chi-square with one degree of freedom = 0.047, p =
0.828).
Let's look at another example, this time looking at the linear relationship between
gender (female) and socio-economic status (ses). The point of this example is that one
(or both) variables may have more than two levels, and that the variables do not have to
have the same number of levels. In this example, female has two levels (male and
female) and ses has three levels (low, medium and high).
a. 0 cells (.0%) have expected count less than 5. The minimum expected count is 21.39.
Again we find that there is no statistically significant relationship between the variables
(chi-square with two degrees of freedom = 4.577, p = 0.101).
Fisher's exact test
The Fisher's exact test is used when you want to conduct a chi-square test but one or
more of your cells have an expected frequency of five or less. Remember that the chisquare test assumes that each cell has an expected frequency of five or more, but the
Fisher's exact test has no such assumption and can be used regardless of how small the
expected frequency is. In SPSS unless you have the SPSS Exact Test Module, you can
only perform a Fisher's exact test on a 2x2 table, and these results are presented by
default. Please see the results from the chi squared example above.
One-way ANOVA
The mean of the dependent variable differs significantly among the levels of program
type. However, we do not know if the difference is between only two of the levels or all
three of the levels. (The F test for the Model is the same as the F test for prog because
prog was the only variable entered into the model. If other variables had also been
entered, the F test for the Model would have been different from prog). To see the
mean of write or each level of program type
From this we can see that the students in the academic program have the highest mean
writing score, while students in the vocational program have the lowest.
Discriminant analysis
Discriminant analysis is used when you have one or more normally distributed interval
independent variables and a categorical dependent variable It is a multivariate
technique that considers the latent dimensions in the independent variables for
predicting group membership in the categorical dependent variable.
Clearly, the SPSS output for this procedure is quite lengthy, and it is beyond the scope of
this page to explain all of it. However, the main point is that two canonical variables are
identified by the analysis, the first of which seems to be more related to program type
than the second.
Factor analysis
Factor analysis is a form of exploratory multivariate analysis that is used to either
reduce the number of variables in a model or to detect relationships among variables.
All variables involved in the factor analysis need to be interval and are assumed to be
normally distributed. The goal of the analysis is to try to identify factors which underlie
the variables. There may be fewer factors than variables, but there may not be more
factors than variables. For our example, let's suppose that we think that there are some
common factors underlying the various test scores. We will include subcommands for
varimax rotation and a plot of the eigenvalues. We will use a principal components
extraction and will retain two factors. (Using these options will make our results
compatible with those from SAS and Stata and are not necessarily the options that you
will want to use).
SPSS
SAS
STATA
Dependent variable
Independent variable
REVIEW QUESTIONS
1. Define SPSS
2. What do you mean by STATA and SAS
- End of Chapter -
LESSON 20
ADVANCED DATA TECHNIQUES
OBJECTIVES
To understand the procedures and applications of following statistical analysis o
o
o
o
Discriminant analysis
ANOVA
Multi dimensional Scaling
Cluster analysis
STRUCTURE
Analysis of variance
Condition for ANOVA
ANOVA model
Discriminant analysis
Factor analysis
Cluster analysis
Ward's method
means can only be used to test differences between two means. When there are more
than two means, it is possible to compare each mean with each other mean using t-tests.
However, conducting multiple t-tests can lead to severe inflation of the Type I error rate.
Analysis of variance can be used to test differences among several means for significance
without increasing the Type I error rate. This chapter covers designs with betweensubject variables. The next chapter covers designs with within-subject variables.
The statistical method for testing the null hypothesis that the means of several
populations are equal is analysis of variance. It uses a single factor, fixed -effects model
to compare the effects of one factor (brands of coffee, varieties if residential housing,
types if retail stores) on a continuous dependent variable. In a fixed effects model, the
levels of the factor are established in advance and the results are not generalizable to
other levels of treatment.
Consider a hypothetical experiment on the effect of the intensity of distracting
background noise on reading comprehension. Subjects were randomly assigned to one
of three groups. Subjects in Group 1 were given 30 minutes to read a story without any
background noise. Subjects in Group 2 read the story with moderate background noise,
and subjects in Group 3 read the story in the presence of loud background noise.
The first question the experimenter was interested in was whether background noise has
any effect at all. That is, whether the null hypothesis: 1 = 2 = 3 is true where 1 is the
population mean for the "no noise" condition, 2 is the population mean for the
"moderate noise" condition, and 3 is the population mean for the "loud noise"
condition. The experimental design therefore has one factor (noise intensity) and this
factor has three levels: no noise, moderate noise, and loud noise.
Analysis of variance can be used to provide a significance test of the null hypothesis that
these three population means are equal. If the test is significant, then the null
hypothesis can be rejected and it can be concluded that background noise has an effect.
In a one-factor between subjects ANOVA, the letter "a" is used to indicate the number of
levels of the factor (a = 3 for the noise intensity example). The number of subjects
assigned to condition 1 is designated as n1; the number of subjects assigned to condition
2 is designated by n2, etc.
If the sample size is the same for all of the treatment groups, then the letter "n" (without
a subscript) is used to indicate the number of subjects in each group. The total number
of subjects across all groups is indicated by "N". If the sample sizes are equal, then N =
(a)(n); otherwise,
N = n1 + n2 + ... + na
Some experiments have more than one between-subjects factor. For instance, consider a
hypothetical experiment in which two age groups (8-year olds and 12-year olds) are
asked to perform a task either with or without distracting background noise. The two
factors are age and distraction.
Assumptions
Analysis of variance assumes normal distributions and homogeneity of variance.
Therefore, in a one-factor ANOVA, it is assumed that each of the populations is
normally distributed with the same variance (2). In between-subjects analyses, it is
assumed that each score is sampled randomly and independently. Research has shown
that ANOVA is "robust" to violations of its assumptions.
This means that the probability values computed in an ANOVA are satisfactorily
accurate even if the assumptions are violated. Moreover, ANOVA tends to be
conservative when its assumptions are violated. This means that although power is
decreased, the probability of a Type I error is as low or lower than it would be if its
assumptions were met. There are exceptions to this rule. For example, a combination of
unequal sample sizes and a violation of the assumption of homogeneity of variance can
lead to an inflated Type I error rate.
Conditions for ANOVA
1. The sample must be randomly selected from normal populations
2. The populations should have equal variances
3. The distance from one value to its group's mean should be independent of the
distances of other values to that mean (independence of error).
4. Minor variations from normality and equal variances are tolerable. Nevertheless, the
analyst should check the assumptions with the diagnostic techniques.
Analysis of variance breaks down or partitions total variability into component parts.
Unlike the 't' test, which uses the sample standard deviations, ANOVA uses squared
deviations of the variance so computation of distances of the individual data points from
their own mean or from the grand mean can be summed.
In ANOVA model, each group has its own mean and values that deviate from that mean.
Similarly, all the data points from all of the groups produce an overall grand mean. The
total deviation is the sum of the squared differences between each data point and the
overall grand mean.
The total deviation of any particular data point may be partitioned into between-groups
variance and within-group variance. The between-group variance represents the effect
of the treatment or factor. The difference of between-groups means imply that each
group was treated differently. The treatment will appear as deviations of the sample
means from the grand mean. Even if this were not so, there would still be some natural
variability among subjects and some variability attributable to sampling. The withingroups variance describes the deviations of the data points within each group from the
sample mean. This results from variability among subjects and from random variation.
It is often called error.
When the variability attributable to the treatment exceeds the variability arising from
error and random fluctuations, the viability of the null hypothesis begins to diminish.
And this is exactly the way the test static for analysis of variance works.
The test statistics for ANOVA is the F ratio. It compares the variance from the last two
sources:
To compute the F ratio, the sum of the squared deviations for the numerator and
denominator are divided by their respective degrees of freedom. By dividing, computing
the variances as an average or mean, thus the term mean square. The number of degrees
of freedom for the numerator, the mean square between groups, is one less than the
number of groups (k-1). The degree of freedom for the denominator, the mean square
within groups, is the total number of observations minus the number of groups (n-k).
If the null hypothesis is true, there should be no difference between the populations, and
the ratio should be close to 1. If the population means are not equal, the numerator
should manifest this difference. The F ratio should be greater than 1. The f distribution
determines the size of ratio necessary to reject the null hypothesis for a particular
sample size and level of significance.
ANOVA model
To illustrate reports one way ANOVA, consider the following hypothetical example. To
find out the number one best business school in India, 20 business magnets were
randomly selected and asked to rate the top 3 B-schools. The ratings are given below:
Data
Discriminant analysis
4. Mapping the results and defining the dimensions - The statistical program
(or a related module) will map the results. The map will plot each product (usually in
two dimensional space). The proximity of products to each other indicates either how
similar they are or how preferred they are, depending on which approach was used. The
dimensions must be labelled by the researcher. This requires subjective judgment and is
often very challenging. The results must be interpreted.
5. Test the results for reliability and validity - Compute R squared to determine
what proportion of variance of the scaled data can be accounted for by the MDS
procedure. An R-square of 0.6 is considered the minimum acceptable level. Other
possible tests are Kruskal's Stress, split data tests, data stability tests (i.e. eliminating
one brand), and test-retest reliability.
Input Data
The input to MDS is a square, symmetric 1-mode matrix indicating relationships among
a set of items. By convention, such matrices are categorized as either similarities or
dissimilarities, which are opposite poles of the same continuum. A matrix is a similarity
matrix if larger numbers indicate more similarity between items, rather than less. A
matrix is a dissimilarity matrix if larger numbers indicate less similarity. The distinction
is somewhat misleading, however, because similarity is not the only relationship among
items that can be measured and analyzed using MDS. Hence, many input matrices are
neither similarities nor dissimilarities.
However, the distinction is still used as a means of indicating whether larger numbers in
the input data should mean that a given pair of items should be placed near each other
on the map, or far apart. Calling the data "similarities" indicates a negative or
descending relationship between input values and corresponding map distances, while
calling the data "dissimilarities" or "distances" indicates a positive or ascending
relationship.
A typical example of an input matrix is the aggregate proximity matrix derived from a
pile-sort task. Each cell xij of such a matrix records the number (or proportion) of
respondents who placed items i and j into the same pile. It is assumed that the number
of respondents placing two items into the same pile is an indicator of the degree to
which they are similar. An MDS map of such data would put items close together which
were often sorted into the same piles.
Another typical example of an input matrix is a matrix of correlations among variables.
Treating these data as similarities (as one normally would), would cause the MDS
program to put variables with high positive correlations near each other, and variables
with strong negative correlations far apart.
Another type of input matrix is a flow matrix. For example, a dataset might consist of
the number of business transactions occurring during a given period between a set of
corporations. Running this data through MDS might reveal clusters of corporations that
whose members trade more heavily with one another than other than with outsiders.
interpretable solutions. Most importantly, however, MDS can be applied to any kind of
distances or similarities, while factor analysis requires us to first compute a correlation
matrix. MDS can be based on subjects' direct assessment of similarities between stimuli,
while factor analysis requires subjects to rate those stimuli on some list of attributes (for
which the factor analysis is performed).
In summary, MDS methods are applicable to a wide variety of research designs because
distance measures can be obtained in any number of ways.
Applications
Marketing
In marketing, MDS is a statistical technique for taking the preferences and perceptions
of respondents and representing them on a visual grid. These grids, called perceptual
maps, are usually two-dimensional, but they can represent more than two.
Potential customers are asked to compare pairs of products and make judgments about
their similarity. Whereas other techniques obtain underlying dimensions from
responses to product attributes identified by the researcher, MDS obtains the
underlying dimensions from respondents' judgments about the similarity of products.
This is an important advantage. It does not depend on researchers' judgments. It does
not require a list of attributes to be shown to the respondents. The underlying
dimensions come from respondents judgments about pairs of products. Because of
these advantages, MDS is the most common technique used in perceptual mapping.
The "beauty" of MDS is that we can analyze any kind of distance or similarity matrix.
These similarities can represent people's ratings of similarities between objects, the
percent agreement between judges, the number of times a subjects fails to discriminate
between stimuli, etc. For example, MDS methods used to be very popular in
psychological research on person perception where similarities between trait descriptors
were analyzed to uncover the underlying dimensionality of people's perceptions of traits
(see, for example Rosenberg, 1977). They are also very popular in marketing research, in
order to detect the number and nature of dimensions underlying the perceptions of
different brands or products & Carmone, 1970).
In general, MDS methods allow the researcher to ask relatively unobtrusive questions
("how similar is brand A to brand B") and to derive from those questions underlying
dimensions without the respondents ever knowing what is the researcher's real interest.
Cluster Analysis
The term cluster analysis (first used by Tryon, 1939) encompasses a number of different
algorithms and methods for grouping objects of similar kind into respective categories.
A general question facing researchers in many areas of inquiry is how to organize
observed data into meaningful structures, that is, to develop taxonomies. In other words
cluster analysis is an exploratory data analysis tool which aims at sorting different
objects into groups in a way that the degree of association between two objects is
maximal if they belong to the same group and minimal otherwise. Given the above,
cluster analysis can be used to discover structures in data without providing an
explanation/interpretation. In other words, cluster analysis simply discovers structures
in data without explaining why they exist.
We deal with clustering in almost every aspect of daily life. For example, a group of
diners sharing the same table in a restaurant may be regarded as a cluster of people. In
food stores items of similar nature, such as different types of meat or vegetables are
displayed in the same or nearby locations. There is countless number of examples in
which clustering plays an important role. For instance, biologists have to organize the
different species of animals before a meaningful description of the differences between
animals is possible. According to the modern system employed in biology, man belongs
to the primates, the mammals, the amniotes, the vertebrates, and the animals. Note how
in this classification, the higher the level of aggregation the less similar are the members
in the respective class. Man has more in common with all other primates (e.g., apes)
than it does with the more "distant" members of the mammals (e.g., dogs), etc. For a
review of the general categories of cluster analysis methods, see Joining (Tree
Clustering), Two-way Joining (Block Clustering), and k-Means Clustering.
Cluster Analysis (CA) is a classification method that is used to arrange a set of cases into
clusters. The aim is to establish a set of clusters such that cases within a cluster are more
similar to each other than they are to cases in other clusters.
Cluster analysis is an exploratory data analysis tool for solving classification problems.
Its object is to sort cases (people, things, events, etc) into groups, or clusters, so that the
degree of association is strong between members of the same cluster and weak between
members of different clusters. Each cluster thus describes, in terms of the data collected,
the class to which its members belong; and this description may be abstracted through
use from the particular to the general class or type.
Cluster analysis is thus a tool of discovery. It may reveal associations and structure in
data which, though not previously evident, nevertheless are sensible and useful once
found. The results of cluster analysis may contribute to the definition of a formal
classification scheme, such as taxonomy for related animals, insects or plants; or suggest
statistical models with which to describe populations; or indicate rules for assigning new
cases to classes for identification and diagnostic purposes; or provide measures of
definition, size and change in what previously were only broad concepts; or find
exemplars to represent classes.
Whatever business you're in, the chances are that sooner or later you will run into a
classification problem. Cluster analysis might provide the methodology to help you
solve.
Procedure for Cluster Analysis
1. Formulate the problem - select the variables that you wish to apply the clustering
technique to
2. Select a distance measure - various ways of computing distance:
Squared Euclidean distance - the square root of the sum of the squared
differences in value for each variable
Manhattan distance - the sum of the absolute differences in value for any variable
Chebychev distance - the maximum absolute difference in values for any variable
Beyond these we can see that (AB) & (C) and (DE) are more similar to each other than to
(FG).
Hence we could construct the following dendrogram (hierarchical classification).
Note that the clusters are joined (fused) at increasing levels of 'dissimilarity'.
The actual measure of dissimilarity will depend upon the method used. It may be a
similarity measure or a distance measure. Distances between points can be calculated by
using an extension of Pythagorus (these are euclidean distances). These measures of
'dissimilarity' can be extended to more than 2 variables (dimensions) without difficulty.
Clustering Algorithms
Having selected how we will measure similarity (the distance measure) we must now
choose the clustering algorithm, i.e. the rules which govern between which points
distances are measured to determine cluster membership. There are many methods
available, the criteria used differ and hence different classifications may be obtained for
the same data. This is important since it tells us that although cluster analysis may
provide an objective method for the clustering of cases there can be subjectivity in the
choice of method. Five algorithms, available within SPSS, are described.
1.
2.
3.
4.
5.
Method Average. SPSS also provides two other methods based on averages,
CENTROID and MEDIAN. Centroid or UPGMC (Unweighted Pair Groups Method
Centroid) uses the group centroid as the average. The centroid is defined as the centre of
a cloud of points. A problem with the centroid method is that some switching and
reversal may take place, for example as the agglomeration proceeds some cases may
need to be switched from their original clusters.
2. Complete Linkage Clustering (Maximum or furthest-neighbour method):
The dissimilarity between two groups is equal to the greatest dissimilarity between a
member of cluster i and a member of cluster j. This method tends to produce very tight
clusters of similar cases.
3. Single Linkage Clustering (Minimum or nearest-neighbour method): The
dissimilarities between two clusters is the minimum dissimilarity between members
of the two clusters. This method produces long chains which form loose, straggly
clusters. This method has been widely used in numerical taxonomy.
SUMMARY
The complete process of generalized hierarchical clustering can be summarized as
follows:
1. Calculate the distance between all initial clusters. In most analyses initial clusters will
be made up of individual cases.
2. Fuse the two most similar clusters and recalculate the distances.
3. Repeat step 2 until all cases are in one cluster.
One of the biggest problems with this Cluster Analysis is identifying the optimum
number of clusters. As the fusion process continues increasingly dissimilar clusters
must be fused, i.e. the classification becomes increasingly artificial. Deciding upon the
optimum number of clusters is largely subjective, although looking at a graph of the
level of similarity at fusion versus number of clusters may help. There will be sudden
jumps in the level of similarity as dissimilar groups are fused.
KEY TERMS
SPSS
Tabulation
Cross-tabulation
ANOVA
Discriminant analysis
Factor analysis
Conjoint analysis
MDS
Cluster analysis
REVIEW QUESTIONS
1. What do you mean by cross-tabulation?
2. Write short notes on statistical packages
3. Explain the step wise procedure for doing Discriminant Analysis.
4. Write short notes on ANOVA.
OBJECTIVES
STRUCTURE
INTRODUCTION
Factor analysis is a general name denoting a class of procedures primarily used for data
reduction and summarization. In marketing research, there may be a large number of
variables, most of which are correlated and which must be reduced to a manageable
level. Relationships among sets of many interrelated variables are examined and
represented in terms of a few underlying factors. For example, store image may be
measured by asking respondents to evaluate stores on a series of items on a semantic
differential scale. The item evaluations may then be analyzed to determine the factors
underlying store image.
In analysis of variance, multiple regression, and discriminant analysis, one variable is
considered as the dependent or criterion variable, and the others as independent or
predictor variables. However, no such distinction is made in factor analysis. Rather,
factor analysis is an interdependence technique in that an entire set of
interdependent relationships is examined.
Factor analysis is used in the following circumstances:
- To identify underlying dimensions, or factors that explain the correlations among a set
of variables. For example, a set of lifestyle statements may be used to measure the
psychographic profiles of consumers. These statements may then be factor analyzed to
identify) the underlying psychographic factors, as illustrated in the department store
example.
- To identify a new, smaller set of uncorrelated variables to replace the original set of
correlated variables in subsequent multivariate analysis (regression or discriminant
analysis). For example, the psychographic factors identified may be used as independent
variables to explaining the differences between loyal and non loyal consumers.
- To identify a smaller set of salient variables from a larger set for use in subsequent
multivariate analysis. For example, a few of the original lifestyle statements that
correlate highly with the identified factors may be used as independent variables to
explain the differences between the loyal and non-loyal users.
Definition
Factor analysis is a class of procedures primarily used for data reduction and
summarization. Factors analysis is an interdependence technique in that an entire set of
interdependent relationships is examined. Factors are defined as an underlying
dimension that explains the correlation among a set of variables
Evolution of Factor Analysis
Charles Spearman first used the factor analysis as a technique of indirect measurement.
When they test human personality and intelligence, a set of questions and tests are
developed for this purpose. They believe that a person gives this set of questions and
tests would respond on the basis of some structure that exists in his mind. Thus, his
responses would form a certain pattern. This approach is based on the assumption that
the underlying structure in answering the questions would be the same in the case of
different respondents.
Even though it is in the field of psychology that factor analysis has its beginning, it has
since been applied to problems in different areas including marketing. Its use has
become far more frequent as a result of the introduction Specialized software packages
such as SPSS, SAS.
Application of Factor Analysis
It can bring out the hidden or latent dimensions relevant in the relationships among
product preferences. Factor analysis is typically used to study a complex product or
service in order to identify the major characteristics (or factors) considered to be
important by consumers of the product or service Example: Researchers for an
automobile (two wheeler) company may ask a large sample of potential buyers to report
(using rating scales) the extent of their agreement or disagreement with the number of
statements such as "A motor bikes breaks are its most crucial part", "Seats should be
comfortable for two members". Researchers apply factor analysis to such a set of data to
identity, which factors such as "safety", "exterior styling", "economy of operations" are
considered important by potential customers. If this information is available, it can be
used to guide the overall characteristics to be designed into the product or to identify
advertising themes that potential buyers aid consider important.
Steps Involved In Conducting the Factor Analysis:
i. Mileage per liter should be high; ii. Maintenance cost should be low; iii. Mileage
should be consistent in all types of roads.
The researcher may then make the judgment that agreement with these set of
statements indicates an underlying concern with the factor of "Economy of operation".
Determine the Method of Factor Analysis
Once it has been determined that factor analysis is an appropriate technique for
analyzing data, an appropriate method must be selected. The approach used to derive
the weights factor score coefficients differentiates the various methods of factor analysis.
The two basic approaches are principal components analysis and common factor
analysis. In principal components analysis, the total variance in the data is considered.
The diagonal of correlation matrix consists of unities, and full variance is brought into
the factor matrix. Principal components analysis is recommended when the
primary concern is to determine the minimum number of factors that will account for
maximum variance in the data for use in subsequent multivariate analysis. The factors
are called principal components.
In common factor analysis, the factors are estimated based only on the common
variance. Communalities are inserted in the diagonal of the correlation matrix. This
method is appropriate when the primary concern is to identify the underlying
dimensions and the common variance is of interest. This method is also known as
principal axis factoring.
Other approaches for estimating the common factors are also available. These include
methods of unweighted least squares, generalized least squares, maximum likelihood,
HA method, and image factoring. These methods are complex and are not
recommended for experienced users.
Determine the Number of Factors
It is possible to compute as many principal components as there are variables, but in
doing so, no parsimony is gained. In order to summarize the information contained in
the original variables, a smaller number of factors should be extracted. The question is,
how many? Several procedures have been suggested for determining the number of
factors. These include a priori determination and approaches based on eigenvalues,
scree plot, percentage variance accounted for, split-half reliability, and significance
tests.
A Prior Determination. Sometimes, because of prior knowledge, the researcher
knows many factors to expect and thus can specify the number of factors to be extracted
beforehand. The extraction of factors ceases when the desired number of factors have
been extracted. Most computer programs allow the user to specify the number of
factors, allowing for an easy implementation of this approach.
This resulted in a set of data in which each of 300 individuals gave a response to each of
100 statements. For any given statement, some individuals were found to agree strongly,
some were found to disagree slightly, some neither agreed nor disagreed with the
statement, and so on. Thus, for each statement, there was a distribution of 300
responses on a 5-point scale.
Three Important Measures
There are three important measures used in the factor analysis:
1. Variance
2. Standardized scores of an individual's responses
3. Correlation coefficient.
1. Variance
A factor analysis is somewhat like regression analysis in that it tries to "best fit" factors
to a scatter diagram of the data in such a way that the factors explain the variance
associated with the responses to each statement.
2. Standardized Scores of an Individual's Responses
To facilitate comparisons of the responses from such different scales, researchers
standardize all of the answers from all of the respondents on all statements and
questions.
one set will all be highly correlated with each other, but they will also be quite
uncorrelated with the responses to the statements in other sets. Since the different sets
of statements are relatively uncorrelated with each other, a separate and distinct factor
relative to motorcycles is associated with each set.
It is already noted that Variance is one of the three important measures used in factor
analysis with standardized responses to each statement used in the study.
Factor analysis selects one factor at a time using procedures that "best fit" each other to
the data. The first factor selected is one that fits the data in such a way that it explains
more of the variance in the entire set of standardized scores than any other possible
factor. Each factor selected after the first factor must be uncorrelated with the factors
already selected. This process continues until the procedures cannot find additional
factors that significantly reduce the unexplained variance in the standardized scores
Output of Factor Analysis
Here, only following six statements (or variables) for simplicity will be used to explain
the output of a factor analysis
XI - Mileage per liter should be high
X2 = Maintenance cost should be low
X3 = Mileage should be consistent in all types of roads.
X4 = Appearance of motorcycle should be masculine
X5 = multiple colors should be available
X6 = Breaks are the most important parts for motorcycles
The results a factor analysis of these six statements will appear in the form shown m the
following table which can be used to illustrate the three important outputs from a factor
analysis and how they can be of use to researchers.
Table 1: Factor Analysis Output of the Motorcycle Study
Factor Loadings
The top six rows of the table are associated with the six statements listed above. The
table shows that the factor analysis has identified three factors (F1, F2 and F3) and the
first three columns are associated with those factors. For example' the first factor can be
written as
F1= 0.86 x 1+ 0.84 x 2 + 0.68 x 3 + 0.10 x 4 + 0.06 x 5 + 0.12 x 6
The 18 numbers located in the six rows and three columns are called as factor loadings
and they are one of the 3 useful outputs obtained from a factor analysis. As shown in
table, each statement has a factor loading associated with a specific factor and a specific
statement is simply the correlation between that factor and that statements
standardized response scores.
Thus, table 1 shows that factor 1 is highly correlated with the responses to statement 1
(0.86 correlation) and also with the response to statement 2 (0.84 correlation).
Table 1 shows that statements 1 and 2 are not highly correlated (0.12 and 0.18
respectively) with factor 2. Thus a factor loading is a measure of how well the factor
"fits" the standardized responses to Statement.
Naming the Factors
From the table, it is clear that factor F1 is good fit on the data from statements 1, 2 and
3, but a poor fit on the other statements. This indicates that statements 1, 2 and 3 are
probably measuring the same basic attitude or value system; and it is this finding that
provides the researchers with evidence that factor exists.
By using the knowledge on the industry and the contents of statements 1, 2, and 3
researchers from Motorcycle Company subjectively concluded from these results that
"Economy of operation" was the factor that tied these statements together in the minds
of the respondents.
Researchers next want to know if the 300 respondents participating in the study mostly
agreed with or disagreed with statements 1, 2 and 3. To answer this question, the
researchers had to look at the 300 standardized responses to each of the statements 1, 2
and 3. They found that the means of these responses to each of statements 1, 2 and 3.
They found that the means of these responses were +0.97, + 1.32, and +1.18,
respectively, for statements 1, 2 and 3, indicating that most respondents agreed with the
three statements as per above discussion on "standardized scores". Since a majority of
respondents had agreed with these statements, the researchers also concluded that the
factor of "economy of operation" was important in the minds of potential motorcycle
customers.
Table also shows that F2 is good fit on statements 4 and 5 but a poor fit on the other
statements. This factor is clearly measuring something different from statements 1, 2, 3
and 6. Factor F3 is good fit only on statement 6 and so it is clearly measuring something
not being statements 1 to 5. Researchers again subjectively concluded that the factor
underlying statements 4 and 5 was "Comfort" and that statement 6 was related to
"safety".
Fit between Data and Factor
Researcher has to find how well all of the identified factors fit the data obtained from all
of the respondents on any given statement. Communalities for each statement indicate
the proportion of the variance in the responses to the statement which is explained by
the three identified factors.
For example, Three factors explain 0.89 (89%) of the variance in all of the responses to
statement 5, but only 0.54 (54%) of the variance in all of the responses to statement 3. It
shows that three factors explain 75% or more of the variance associated with statements
1, 2, 4, 5 and 6, but only about half of statement 3's variance .Researchers can use these
communalities to make a judgment about for most of the variance associated with each
of the six statements in this example, the three factors fit the data quite well.
How well any given factor fits the data from all of the respondents on all of the
statements can be determined by eigenvalue. There is an eigenvalue associated with
each of the factors.
When a factor's eigenvalue is divided by the number of statements used in factor
analysis the resulting figure is the proportion of the variance in the entire set of
standardized response scores which is explained by that factor i.e. each eigenvalue is
divided by 6 the number of statements. For example factor F1 explains 0.3226 (32%)
of the variance of the standardized response scores from all of the respondents on all six
statements. By adding these figures for the three factors, three factors together explain
Factor analysis
Communalities
Factor loading
Correlation matrix
Eigen value
IMPORTANT QUESTIONS
1. What are the applications of factor analysis?
2. What is the significance of factor loading in factor analysis?
3. What do you mean by Eigen value?
STRUCTURE
Conjoint analysis
Basics of conjoint analysis
Steps involved in conjoint analysis
Application of conjoint analysis
their desirability. Conjoint procedures attempt to assign values to the levels of each
attribute, so that the resulting values or utilities attached to the stimuli match, as closely
as possible the input evaluations provided by the respondents. The underlying
assumption is that any set of stimuli, such as products, brands, or stores, is evaluated as
a bundle of attributes.
Conjoint Analysis is a technique that attempts to determine the relative importance
consumers attach to salient attributes and the utilities they attach to the levels of
attributes.
In a real purchase situation, however, consumers do not make choices based on a single
attribute like comfort. Consumers examine a range of features or attributes and then
make judgments or trade-offs to determine their final purchase choice. Conjoint analysis
examines these trade-offs to determine the combination of attributes that will be most
satisfying to the consumer, in other words, by using conjoint analysis a company can
determine the optimal features for their product or service. In addition, conjoint
analysis will identify the best advertising message by identifying the features that are
most important in product choice.
Like multidimensional scaling, conjoint analysis relies on respondents' subjective
evaluations. However, in MDS the stimuli are products or brands. In conjoint analysis,
the stimuli are combinations of attribute levels determined by the researcher. The goal
in MDS is to develop a spatial map depicting the stimuli in a multidimensional
perceptual or preference space. Conjoint analysis, on the other hand, seeks to develop
the part-worth or utility functions describing the utility consumers attach to the levels of
each attribute. The two techniques are complementary.
In sum, the value of conjoint analysis is that it predicts what products or services people
will choose and assesses the weight people give to various factors that underlie their
decisions. As such, it is one of the most powerful, versatile and strategically important
research techniques available.
Statistics and Terms Associated with Conjoint Analysis
The important statistics and terms associated with conjoint analysis include:
Part-worth functions. Also called utility functions, these describe the utility
consumers attach to the levels of each attribute.
Relative importance weights. Indicate which attributes are important in
influencing consumer choice. These weights are estimated.
Attribute levels. Denote the values assumed by the attributes.
Full profiles. Full profiles or complete profiles of brands are constructed in
terms of all the attributes by using the attribute levels specified by the design.
Pair-wise tables. The respondents evaluate two attributes at a time until all
the required pairs of attributes have been evaluated.
Cyclical designs. Designs employed to reduce the number of paired
comparisons.
Any number of algorithms may be used to estimate utility functions. The original
methods were monotonic analysis of variance or linear programming techniques, but
these are largely obsolete in contemporary marketing research practice. Far more
popular are the Hierarchical Bayesian procedures that operate on choice data. These
utility functions indicate the perceived value of the feature and how sensitive consumer
perceptions and preferences are to changes in product features.
A Practical Example of Conjoint Analysis
Conjoint analysis presents choice alternatives between products/services defined by sets
of attributes. This is illustrated by the following choice: would you prefer a flight with
regular seats, that costs Rs.400 and takes 5 hours, or a flight which costs Rs.700, has
extra-wide seats and takes 3 hours?
Extending this, we see that if seat comfort, price and duration are the only relevant
attributes, there are potentially eight flight choices.
Given the above alternatives, product 4 is very likely the most preferred choice, while
product 5 is probably the least preferred product. The preference for the other choices is
determined by what is important to that individual.
Conjoint analysis can be used to determine the relative importance of each attribute,
attribute level, and combinations of attributes. If the most preferable product is not
feasible for some reason (perhaps the airline simply cannot provide extra-wide seats and
a 3 hour arrival time at a price of Rs400) then the conjoint analysis will identify the next
most preferred alternative. If you have other information on travelers, such as
background demographics, you might be able to identify market segments for which
distinct products may be appealing. For example, the business traveller and the vacation
traveller may have very different preferences which could be met by distinct flight
offerings.
You can now see the value of conjoint analysis. Conjoint analysis allows the researcher
to examine the trade-offs that people make in purchasing a product. This allows the
researcher to design products/services that will be most appealing to a specific market.
In addition, because conjoint analysis identifies important attributes, it can be used to
create advertising messages that will be most persuasive.
In evaluating products, consumers will always make trade-offs. A traveller may like the
comfort and arrival time of a particular flight, but reject purchase due to the cost. In this
case, cost has a high utility value. Utility can be defined as a number which represents
the value that consumers place on an attribute. In other words, it represents the relative
"worth" of the attribute. A low utility indicates less value; a high utility indicates more
value.
The following figure presents a list of hypothetical utilities for an individual consumer:
This consumer places a greater value on a 3 hour flight (the utility is 42) than on
a 5 hour flight (utility is 22).
This consumer does not differ much in the value that he or she places on comfort.
That is, the utilities are quite close (12 vs.15).
This consumer places a much higher value on a price of Rs.400 than a price of
Rs.700.
The preceding example depicts an individual's utilities. Average utilities can be
calculated for all consumers or for specific subgroups of consumers.
These utilities also tell us the extent to which each of these attributes drives the decision
to choose a particular flight. The importance of an attribute can be calculated by
examining the range of utilities (that is, the difference between the lowest and highest
utilities) across all levels of the attribute. That range represents the maximum impact
that the attribute can contribute to a product.
Using the hypothetical utilities presented earlier, we can calculate the relative
importance of each of the three attributes. The range for each attribute is given below:
These ranges tell us the relative importance of each attribute. Cost is the most important
factor in product purchase as it has the highest range of utility values. Cost is followed in
importance by the duration of the flight. Based on the range and value of the utilities, we
can see that seat comfort is relatively unimportant to this consumer. Therefore,
advertising which emphasizes seat comfort would be ineffective. This person will make
his or her purchase choice based mainly on cost and then on the duration of the flight.
Marketers can use the information from utility values to design products and/or services
which come closest to satisfying important consumer segments. Conjoint analysis will
identify the relative contributions of each feature to the choice process. This technique,
therefore, can be used to identify market opportunities by exploring the potential of
product feature combinations that are not currently available.
Choice Simulations
In addition to providing information on the importance of product features, conjoint
analysis provides the opportunity to conduct computer choice simulations. Choice
simulations reveal consumer preference for specific products defined by the researcher.
In this case, simulations will identify successful and unsuccessful flight packages before
they are introduced to the market!
For example, let's say that the researcher defined three flights as follows:
The conjoint simulation will indicate the percentage of consumers that prefer each of
the three flights. The simulation might show that consumers are willing to travel longer
if they can pay less and are provided a meal. Simulations allow the researcher to
estimate preference, sales and share for new flights before they come to market.
Readers might be worried at this point about the total number of cards that need to be
rated by a single respondent. Fortunately, we are able to use statistical manipulations to
cut down on the number of cards. In a typical conjoint study, respondents only need to
rate between 10-20 cards.
This data would be input to the conjoint analysis. Utilities can then be calculated and
simulations can be performed to identify which products will be successful and which
should be changed. Price simulations can also be conducted to determine sensitivity of
the consumer to changes in prices.
A wide variety of companies and service organizations have successfully used conjoint
analysis
A conjoint analysis was developed using a number of attributes such as saving on energy
bills, efficiency rating of equipment, safety record of energy source, and dependability of
energy source. The conjoint analysis identified that cost savings and efficiency were the
main reasons for converting appliances to gas. The third most important reason was
cleanliness of energy source. This information was used in marketing campaigns in
order to have the greatest effect.
A natural gas utility used conjoint analysis to evaluate which advertising message would
be most effective in convincing consumers to switch from other energies to natural gas.
Previous research failed to discover customer's specific priorities - it appeared that the
trade-offs that people made were quite subtle.
Advantages
Disadvantages
Only a limited set of features can be used because the number of combinations
increases very quickly as more features are added.
Information gathering stage is complex.
Difficult to use for product positioning research because there is no procedure for
converting perceptions about actual features to perceptions about a reduced set
of underlying features
Respondents are unable to articulate attitudes toward new categories
Applications of conjoint analysis have been made in consumer goods, industrial goods,
financial and other services. Moreover, these applications have spanned all areas of
marketing. A recent survey of conjoint analysis reported applications in the areas of new
product/concept identification, competitive analysis, pricing, market segmentation,
advertising, and distribution.
SUMMARY
This chapter has given over view conjoint analysis in detail. Conjoint analysis, also
called multi-attribute compositional models. Today it is used in many of the social
sciences and applied sciences including marketing, product management, and
operations research.
KEY TERMS
IMPORTANT QUESTIONS
1. What are the applications of conjoint analysis?
2. Explain the procedure of performing Conjoint Analysis with one practical example.
REFERENCE BOOKS
1. Robert Ferber, Marketing Research, New York: McGraw Hill, Inc. 1948
2. Dennis, Child, The Essentials of Factor Analysis, New York 1973
3. Cooley, William, W., and Lohnes, Paul R., Multivariate Data Analysis, New York:
John Wiley and Sons. 1971.
- End of Chapter -
LESSON 23
APPLICATION OF RESEARCH TOOLS
OBJECTIVES
o
o
o
STRUCTURE
INTRODUCTION
Research Methodology is becoming a necessary tool for its application of all functional
areas of management such as Marketing, Human Resources and Operations
Management. There is an increasing realisation of the importance of research
methodology in various quarters. This is reflected in the increasing use of research
methodology in various domains of management. Here, a brief description of typical
application of research methodology is given below:
APPLICATIONS OF MARKETING RESEARCH
Strategic
Tactical
As the list shows, marketing research tackles a wide variety of subjects. The list is only
indicative, and the applications of marketing research in reality can be useful for almost
any major decision related to marketing. The next sections discuss some typical
application areas.
Concept Research
During a new product launch, there would be several stages-for example, concept
development, concept testing, prototype development and testing, test marketing in a
designated city or region, estimation of total market size based on the test marketing,
and then a national rollout or withdrawal of the product based on the results.
The first stage is the development of a concept and its testing. The concept for a new
product may come from several sources-the idea may be from a brain storming session
consisting of company employees, a focus group conducted among consumers, or the
brainwave of a top executive. Whatever may be its source, it is generally researched
further through what is termed as concept testing, before it goes into prototype or
product development stages.
A concept test takes the form of developing a description of the product, its benefits,
how to use it, and so on, in about a paragraph, and then asking potential consumers to
rate how much they like the concept, how much they would be willing to pay for the
product if introduced, and similar questions. As an example, the concept statement for a
fabric softener may read as follows.
This fabric softener cum whitener is to be added to the wash cycle in a machine or to the
bucket of detergent in which clothes are soaked. Only a few drops of this liquid will be
needed per wash to whiten white clothes and also soften them by eliminating static
charge. It will be particularly useful for woolens, undergarments and babys or childrens
clothes. It will have a fresh fragrance, and will be sold in handy 200 ml, bottles to last
about a month. It can also replace all existing blues with the added benefit of a softener.
This statement can be used to survey existing customers of 'blues' and whiteners, and we
could ask customers for their reactions on pack size, pricing, colour of the liquid, ease of
use, and whether or not they would buy such a product. More complex concept tests can
be done using Conjoint Analysis where specific levels of price or product/service
features to be offered are pre-determined and reactions of consumers are in the form of
ratings given to each product concept combining various features. This is then used to
make predictions about which product concepts would provide the highest utility to the
consumer, and to estimate market shares of each concept. The technique of Conjoint
Analysis is discussed with an example in Part II of the book.
Product Research
Apart from product concepts, research helps to identify which alternative packaging is
most preferred, or what drives a consumer to buy a brand or product category itself, and
specifics of satisfaction or dissatisfaction with elements of a product. These days, service
elements are as important as product features, because competition is bringing most
products on par with each other.
An interesting area of research into pricing has been determining price elasticity at
various price points for a given brand through experiments or simulations. Price
framing, or what the consumer compares (frames) price against, is another area of
research. For example, one consumer may compare the price of a car against an
expensive two-wheeler (his frame of reference), whereas another may compare it with
an investment in the stock market or real estate. Another example might be the interest
earned from a fixed deposit, which serves as a benchmark for one person before he
decides to invest in a mutual fund, whereas for another, the investment may be a
substitute for buying gold, which earns no interest. In many cases, therefore, it is the
frame of reference used by the customer which determines 'value' for him of a given
product. There are tangible as well as intangible (and sometimes not discernible)
aspects to a consumer's evaluation of price. Some of the case studies at the end of Part I
include pricing or price-related issues as part of the case.
Distribution Research
Traditionally, most marketing research focuses on consumers or buyers. Sometimes this
extends to potential buyers or those who were buyers but have switched to other brands.
But right now, there is a renewed interest in the entire area of logistics, supply chain,
and customer service at dealer locations. There is also increasing standardisation from
the point of view of brand building, in displays at the retail level, promotions done at the
distribution points. Distribution research focuses on various issues related to the
distribution of products including service levels provided by current channels, frequency
of sales persons visits to distribution points, routing/transport related issues for
deliveries to and from distribution points throughout the channel, testing of new
channels, channel displays, linkages between displays and sales performance, and so on.
As an example, a biscuit manufacturer wanted to know how it could increase sales of a
particular brand of biscuits in cinema theatres. Should it use existing concessionaires
selling assorted goods in theatres, or work out some exclusive arrangements? Similarly,
a soft drink manufacturer may want to know where to set up vending machines.
Potential sites could include roadside stalls, shopping malls, educational institutions,
and cinema theatres. Research would help identify factors that would make a particular
location a success.
In many service businesses where a customer has to visit the location, it becomes very
important to research the location itself. For example, a big hotel or a specialty
restaurant may want to know where to locate themselves for better visibility and
occupancy rates. Distribution research helps answer many of these questions and
thereby make better marketing decisions.
Advertising Research
The two major categories of research in advertising are:
1. Copy
2. Media
Copy Testing
This is a broad term that includes research into all aspects of advertising-brand
awareness, brand recall, copy recall (at various time periods such as day after recall,
week after recall), recall of different parts of the advertisement such as the headline for
print ads, slogan or jingle for TV ads, the star in an endorsement and so on. Other
applications include testing alternative ad copies (copy is the name given to text or
words used in the advertisement, and the person in the advertising agency responsible
for writing the words is known as the copy writer) :or a single ad, alternative layouts (a
layout is the way all the elements of, the advertisement are laid out in a print
advertisement) with the same copy, testing of concepts or storyboards (a storyboard is a
scene-by-scene drawing of a TV commercial which is like a rough version before the ad
is actually shot on film) of TV commercials to test for positive/negative reactions, and
many others. Some of these applications appear in our discussion of Analysis of
Variance (ANOVA) in Part II and some case studies elsewhere in the book.
A particular class of advertising research is known as Tracking Studies. When an
advertising campaign is running, periodic sample surveys known as tracking studies can
be conducted to evaluate the effect of the campaign over a long period of time such as
six months or one year, or even longer. This may allow marketers to alter the advertising
theme, content, media selection or frequency of airing / releasing advertisements and
evaluate the effects. As opposed to a snapshot provided by a one-time survey, tracking
studies may provide a continuous or near-continuous monitoring mechanism. But here,
one should be careful in assessing the impact of the advertising on sales, because other
factors could change along with time. For example, the marketing programmes of the
sponsor and the competitors could vary over time. The impact on sales could be due to
the combined effect of several factors.
Media Research
The major activity under this category is research into viewership of specific television
programmes on various TV channels. There are specialised agencies like A.C. Nielsen
worldwide which offer viewership data on a syndicated basis (i.e., to anyone who wants
to buy the data). In India, both ORG-MARG and IMRB offer this service. They provide
people meter data with brand names of TAM and INTAM which is used by advertising
agencies when they draw up media plans for their clients. Research could also focus on
print media and their readership. Here again, readership surveys such as the National
Readership Survey (NRS) and the Indian Readership Survey (IRS) provide syndicated
readership data. These surveys are now conducted almost on a continuous basis in India
and are helpful to find out circulation and readership figures of major print media. ABC
(Audit Bureau of Circulations) is an autonomous body which provides audited figures on
the paid circulation (number of copies printed and sold) of each newspaper and
magazine, which is a member of ABC.
Media research can also focus on demographic details of people reached by each
medium, and also attempt to correlate consumption habits of these groups with their
media preferences. Advertising research is used at all stages of advertising, from
information to the company and enable it to make a more rational and effective effort in
maximizing its return from sales.
THE CONCEPT OF MARKET POTENTIAL
Market potential has been defined as "the maximum demand response possible for a
given group of customers within a well-defined geographic area for a given product or
service over a specified period of time under well-defined competitive and
environmental conditions".
We will elaborate this comprehensive definition. First, market potential is the maximum
demand response under certain assumptions. It denotes a meaningful boundary
condition on ultimate demand.
Another condition on which the concept of market potential depends is a set of relevant
consumers of the product or service. It is not merely the present consumer who is to be
included but also the potential consumer as maximum possible demand is to be
achieved. Market potential will vary depending on which particular group of consumers
is of interest.
Further, the geographic area for which market potential is to be determined should be
well-defined. It should be divided into mutually exclusive subsets of consumers so that
the management can assign a sales force and supervise and control the activities in
different territories without much difficulty.
Another relevant aspect in understanding the concept of market potential is to clearly
know the product or service for which market potential is to be estimated. Especially in
those cases where the product in question can be substituted by another, it is desirable
to have market potential for the product class rather than that particular product. For
example, tea is subjected to a high degree of cross-elasticity of demand with coffee.
It is necessary to specify the time period for which market potential is to be estimated.
The time period should be so chosen that it coincides with planning periods in a firm.
Both short and long-time periods can be used depending on the requirements of the
firm.
Finally, a clear understanding of environmental and competitive conditions relevant in
case of a particular product or service is necessary if market potential is to be useful.
What is likely to be the external environment? What is likely to be the nature and extent
of competition? These are relevant questions in the context of any estimate of market
potential since these are the factors over which the firm has no control.
It may be emphasised that market potential is not the same thing as sales potential and
sales forecast. It is only when "a market is saturated can the industry sales forecast be
considered equivalent to market potential". Such a condition is possible in case of well
established and mature products. Generally, the industry sales forecast will be less than
the market potential. Likewise, a company's sales forecast will be less than its sales
potential. The former is a point estimate of the future sales, while the latter represents a
boundary condition which the sales might reach in an ideal situation. "In the latter
sense, sales potential is to a firm what market potential is to an industry or product
class: both represent maximum demand response and are boundary conditions".
Brand Positioning
Brand positioning is a relatively new concept in marketing. The concept owes its origin
to the idea that each brand occupies a particular space in the consumer's mind,
signifying his perception of the brand in question in relation to other brands. While
product or brand positioning has been defined by various authors in different ways, the
underlying meaning conveyed through these definitions seems to be the same. Instead
of giving several definitions, we may give one here. According to Green and Tull,
"Brand positioning and market segmentation appear to be the hallmarks of today's
marketing research. Brand (or service) positioning deals with measuring the perceptions
that buyers hold about alternative offerings".
From this definition it is evident that the term 'position' reflects the essence of a brand
as perceived by the target consumer in relation to other brands. In view of this the
management's ability to position its product or brand appropriately in the market can be
a major source of company's profits. This seems to be an important reason for the
emergence of product or brand positioning as a major area in marketing research.
Components of Positioning
Positioning comprises four components. The first component is the product class or the
structure of the market in which a company's brand will compete. The second
component is consumer segmentation. One cannot think of positioning a brand without
considering the segment m which it is to be offered. Positioning and segmentation are
inseparable. The third component is the consumers perception of the company's brand
in relation to those of the competitors. Perceptual mapping is the device by which the
company can know this. Finally, the fourth component positioning is the benefit offered
by the company's brand. A consumer can allot a position in his mind to a brand only
when it is beneficial to him. The benefits may be expressed as attributes or dimensions
in a chart where brands are fitted to indicate the consumer's perceptions.
As perceptual maps are used to indicate brand positioning, blank spaces in such maps
show that a company can position its brand in one or more of such spaces.
TECHNIQUES FOR PERCEPTUAL MAPPING
There are a number of techniques for measuring product positioning. Some of these
which are important are:
Factor analysis
Cluster analysis Multi-dimensional scaling.
We will not go into the detailed mechanism of these techniques. All the same, we will
briefly explain the techniques.
Image profile analysis
This technique is the oldest and most frequently used for measuring the consumers
perceptions of competitive brands or services. Normally, a 5 or 7 point numerical scale
used. A number of functional and psychological attributes are selected. The respondent
is asked to show his perception of each brand in respect of each attribute on the 5 or 7
point scale.
It will be seen that the figures provides some insight as to which brands are competing
with each other and on what attribute(s). This technique has some limitations. First, if
the number of brands is large, it may not be possible to plot all the brands in a single
figure. Second, there is an implicit assumption in this technique that all attributes are
equally important and independent of each other. This is usually not true. However, this
limitation can be overcome by using the technique of factor analysis.
Factor analysis
As regards factor analysis, it may be pointed out that its main object is to reduce a large
number of variables into a small number of factors or dimensions. In Chapter 17, two
examples have been given to illustrate the use of factor analysis. The discussion also
brings out some major limitations of the method.
Cluster analysis
Cluster analysis is used to classify consumers or objects into a small number of mutually
exclusive and exhaustive groups. With, the help of cluster analysis it is possible to
separate brands into clusters or groups so that the brand within a cluster is similar to
other brands belonging to the same cluster and is very different from brands included in
other clusters. This method has been discussed in Chapter
Multi-dimensional scaling
Multi-dimensional scaling too has been discussed in Chapter 17, pointing out how
perceptual maps can be developed on the basis of responses from consumers. In this
connection, two illustrations of perceptual maps were given. The first illustration related
to select Business Schools based on hypothetical data On the basis of two criteria, viz.
how prestigious and quantitative an MBA course is different Business Schools have been
shown in the map. It will be seen that the MBA course of Business School 'C is extremely
different from that offered by Business School Q Points which are close to each other
indicate similarity of the MBA courses in the students perception. The second
illustration related to four brands of washing soaps based on a survey data from
Calcutta. This is a non-attribute based example where a paired comparison for four
high- and-medium-priced detergents Surf, Sunlight, Gnat and Key was undertaken. As
mentioned there, Sunlight and Surf are closest and Surf and Key are farthest. In other
words the first two brands are most similar and the remaining two are most dissimilar.
How the points in the figures for the four brands have been arrived at has been
explained at length in that chapter and so is not repeated here.
Subroto Sengupta has discussed at length product positioning in his book. While
explaining different techniques of product positioning, he has shown how the concept of
positioning can be used to improve the image of the concerned product or brand. He has
given a number of examples covering a wide variety of products such as coffee, soft
drinks, washing soaps, toilet soaps, shampoos and magazines As Sengupta points out
the perceptual maps of product class also indicate holes or vacant positions in the
market. These open spaces can be helpful to the management in suggesting new product
opportunities as also possibilities for repositioning of old products. While it is true that
the management does get the clues on preferred attributes of the product in question, it
is unable to know all the relevant features of the new product such as its form, package
and price. This problem can be overcome through the application of the conjoint
analysis. In addition, Sengupta has discussed some research studies in respect of
advertising positioning.
We now give a detailed version of a study indicating how a brand which was putting up a
poor performance in the market was repositioned. As a result, it improved its image and
contributed to increased market share and profits.
When to do Marketing Research?
Marketing research can be done when:
There is an information gap which can be filled by doing research.
The cost of filling the gap through marketing research is less than the cost of
taking a wrong decision without doing the research.
o The time taken for the research does not delay decision-making beyond
reasonable limits.
o A delay can have many undesirable effects, like competitors becoming aware of
strategies or tactics being. Contemplated, consumer opinion changing between
the beginning and end of the study, and so forth.
o
o
study was that there was no demand, and that the camera would not succeed, if
launched. The company went ahead and launched it anyway, and it was a huge success.
The camera was Hot Shot. It was able to tap into the need of consumers at that time for
an easy-to-use camera at an affordable price.
Thus marketing research is not always the best or only source of information to be used
for making decisions. It works best when combined with judgment, intuition,
experience, and passion. For instance, even if marketing research were to show there
was demand for a certain type of product, it still depends on the design and
implementation of the appropriate marketing plans to make it succeed. Further,
competitors could take actions which were not foreseen when marketing research was
undertaken. This also leads us to conclude that the time taken for research should be the
minimum possible, if we expect the conditions to be dynamic, or fast changing.
Differences in Methodology
The reader may be familiar with research studies or opinion polls conducted by different
agencies showing different results. One of the reasons why results differ is because the
methodology followed by each agency is usually different. The sampling method used,
the sample size itself, the representativeness of the population, the quality of field staff
who conduct interviews, and conceptual skills in design and interpretation all differ
from agency to agency. Minor differences are to be expected in sample surveys done by
different people, hut major differences should be examined for the cause, which will
usually lead us to the different methodologies adopted by them. Based on the credibility
of the agency doing the research and the appropriateness of the methodology followed,
the user decides which result to rely upon. A judgment of which methodology is more
appropriate for the research on hand comes from experience of doing a variety of
research.
To summarise, it is important to understand the limitations of marketing research, and
to use it in such a way that we minimise its limitations.
Complementary Inputs for Decision-Making
Along with marketing research, marketing managers may need to look into other
information while making a decision. For example, our corporate policy may dictate that
a premium image must be maintained in all activities of our company. On the other
hand, marketing research may tell us that consumers want a value-for-money product.
This creates a dilemma for the basic corporate policy, which has to be balanced with
consumer perception as measured by marketing research.
Other inputs for decision-making could be growth strategies for the brand or product,
competitors' strategies, and regulatory moves by the government and others. Some of
these are available internally-for example, corporate policy and growth plans may be
documented internally. Some other inputs may come from a marketing intelligence cell
if the company has one. In any case, marketing decisions would be based on many of
these complementary inputs, and not on the marketing research results alone.
The methods of forecasting can be divided into two broad categories, viz. subjective or
qualitative methods and objective or quantitative methods. These can be further divided
into several methods. Each of these methods is discussed below.
I. Subjective / Qualitative Methods
There are four subjective methods - field sales force, jury of executives, users'
expectations and delphi. These are discussed here briefly.
a. Field sales force
Some companies ask their salesmen to indicate the most likely sales for a specified
period in the future. Usually the salesman is asked to indicate anticipated sales for each
account in his territory. These forecasts are checked by district managers who forward
them to the company's head office. Different territory-forecasts are then combined into
a composite forecast at the head office. This method is more suitable when a short-term
forecast is to be made as there would be no major changes in this short period affecting
the forecast. Another advantage of this method is that it involves the entire sales force
which realises its responsibility to achieve the target it has set for itself. A major
limitation of this method is that sales force would not take an overall or broad
perspective and hence may overlook some vital factors influencing the sales. Another
limitation is that salesmen may give somewhat low figures in their forecasts thinking
that it may be easier for them to achieve those targets. However, this can be offset to a
certain extent by district managers who are supposed to check the forecasts.
b. Jury of executives
Some companies prefer to assign the task of sales forecasting to executives instead of a
sales force. Given this task each executive makes his forecast for the next period. Since
each has his own assessment of the environment and other relevant factors, one forecast
is likely to be different from the other. In view of this it becomes necessary to have an
average of these varying forecasts. Alternatively, steps should be taken to narrow down
the difference in the forecasts. Sometimes this is done by organising a discussion
between the executives so that they can arrive at a common forecast. In case this is not
possible, the chief executive may have to decide which of these forecasts is acceptable as
a representative one.
This method is simple. At the same time, it is based on a number of different viewpoints
as opinions of different executives are sought. One major limitation of this method is
that the executives' opinions are likely to be influenced in one direction on the basis of
general business conditions.
c. Users' expectations
Forecasts can be based on users' expectations or intentions to purchase goods and
services. It is difficult to use this method when the number of users is large. Another
limitation of this method is that though it indicates users' 'intentions' to buy, the actual
purchases may be far less at a subsequent period. It is most suitable when the number of
buyers is small such as in case of industrial products.
d. The Delphi method
This method too is based on the experts' opinions. Here, each expert has access to the
same information that is available. A feedback system generally keeps them informed of
each others' forecasts but no majority opinion is disclosed to them. However, the experts
are not brought together. This is to ensure that one or more vocal experts do not
dominate other experts.
The experts are given an opportunity to compare their own previous forecasts with those
of the others and revise them. After three or four rounds, the group of experts arrives at
a final forecast.
The method may involve a large number of experts and this may delay the forecast
considerably. Generally it involves a small number of participants ranging from 10 to
40.
It will be seen that both the jury of executive opinion and the Delphi method are based
on a group of experts. They differ in that in the former, the group of experts meet,
discuss the forecasts, and try To arrive at a commonly agreed forecast while in the latter
the group of experts never meet. As mentioned earlier, this is to ensure that no one
person dominates the discussion thus influencing the forecast. In other words, the
Delphi method retains the wisdom of a group and at the same time reduces the effect of
group pressure. An approach of this type is more appropriate when long-term forecasts
are involved.
In the subjective methods, judgments are an important ingredient. Before attempting a
forecast, the basic assumptions regarding environmental conditions as also competitive
behaviour must be provided people involved in forecasting. An important advantage of
subjective methods is that they are easily understood. Another advantage is that the cost
involved in forecasting is quite low.
As against these advantages, subjective methods have certain limitations also. One
major limitation is the varying perceptions of people involved in forecasting. As a result,
wide variance is found in forecasts. Subjective methods are suitable when forecasts are
to be made for highly technical products which have a limited number of customers.
Generally, such methods are used for industrial products. Also when cost of forecasting
is to be kept to a minimum, subjective methods may be more suitable.
II. Objective / Quantitative or Statistical Methods
Based on statistical analysis, these methods enable the researcher to make forecasts on a
more objective basis. It is difficult to make a wholly accurate forecast for there is always
an element of uncertainty regarding the future'. Even so, statistical methods are likely to
be more useful as they are more scientific and hence more objective.
a. Time Series
In time-series forecasting, the past sales data are extrapolated as a linear or a curvilinear
trend. Even if such data are plotted on a graph, one can extrapolate for the desired time
period. Extrapolation can be made with the help of statistical techniques.
It may be noted that time-series forecasting is most suitable to stable situations where
the future ids will largely be an extension of the past. Further, the past sales data should
have distinctive ids from the random error component for a time-series forecasting to be
suitable.
Before using the time-series forecasting one has to decide how far back in the past one
can go. It may be desirable to use the more recent data as conditions might have been
different in the remote past. Another issue pertains to weighting of time-series data. In
other words, should equal weight be given to each time period or should greater
weightage be given to more recent data? Finally, should data be decomposed into
different components, viz. trend, cycle, season and error? We now discuss methods, viz.
moving averages, exponential smoothing and decomposition of time series.
b. Moving average
This method uses the last 'n' data points to compute a series of average in such a way
that each time latest figure is used and the earliest one dropped. For example, when we
have to calculate a five monthly moving average, we first calculate the average of
January, February, March, April and May by adding the figures of these months, and
dividing the sum by five. This will give one figure. In the next calculation, the figure for
June will be included and that for January dropped thus giving a new average. Thus a
series of averages is computed. The method is called as 'moving' average as it uses w
data point each time and drops the earliest one.
In a short-term forecast, the random fluctuations in the data are of major concern. One
method of minimizing the influence of random error is to use an average of several past
data points. This is achieved by the moving average method. It may be noted that in a
12-month moving average, the effect of seasonality is removed from the forecast as data
points for every season are included before computing the moving average.
c. Exponential smoothing
A method which has been receiving increasing attention in recent years is known as
exponential smoothing. It is a type of moving average that 'smoothens' the time-series.
When a large number of forecasts are to be made for a number of items, exponential
smoothing is particularly suitable as it combines the advantages of simplicity of
computation and flexibility.
d. Time-series decomposition
This method consists of measuring the four components of a time-series (i) trend, (ii)
cycle, (iii) season, and (iv) erratic movement.
(i) The trend component indicates long-term effects on sales that are caused by
such factors as income, population, industrialisation and technology. The time
period a trend function varies considerably from product to product However, it
is usually taken as any period in excess of the time period required for a business
cycle (which averages at 4-5 years).
(ii) The cyclical component indicates some sort of a periodicity in the general
economic activity. When data are plotted, they yield a curve with peaks and
troughs, indicating increases and falls in the trend series with a certain
periodicity. A careful study of the impact of a business cycle must be made on the
sale of each product. Cyclical forecasts are likely to be more accurate for the longterm than for the short term.
(iii) The seasonal component reflects changes in sales levels due to factors such as
weather, festivals, holidays, etc. There is a consistent pattern of sales for period
within a year.
(iv) Finally, the erratic movements in data arise on account of events such as
strikes, lockouts, price wars, etc. The decomposition of time-series enables
identification of the error component from the trend, cycle and season which are
systematic components.
Casual or Explanatory Methods
Causal or explanatory methods are regarded as the most sophisticated methods of
forecasting sales. These methods yield realistic forecasts provided relevant data are
available on the- major variables influencing changes in sales. There are three distinct
advantages of causal methods. First, turning points in sales can be predicted more
accurately by these methods than by time-series methods. Second, the use of these
methods reduces the magnitude of the random component far more than it may be
possible with the time-series methods. Third, the use of such methods provides greater
insight to causal relationships. This facilitates the management in marketing decision
making. Isolated sales recasts on the basis of time-series methods would not be helpful
in this regard.
Causal methods can be either (i) leading indicators or (ii) regression models. These are
briefly discussed here.
(i) Leading indicators
Sometimes one finds that changes in sales of a particular product or service are
preceded by changes one or more leading indicators. In such cases, it is necessary to
identify leading indicators and to closely observe changes in them. One example of a
leading indicator is the demand for various household appliances which follows the
construction of new houses. Likewise, the demand for many durables is preceded by an
increase in disposable income. Yet another example is of number of births. The demand
for baby food and other goods needed by infants can be ascertained by the number of
births in a territory. It may be possible to include leading indicators in regression
models.
(ii) Regression models
Linear regression analysis is perhaps the most frequently used and the most powerful
method among casual methods. As we have discussed regression analysis in detail in the
preceding chapters on Bivariate Analysis and Multivariate Analysis, we shall only dwell
on a few relevant points.
First, regression models indicate linear relationships within the range of observations
and at the les when they were made. For example, if a regression analysis of sales is
attempted on the basis of independent variables of population sizes of 15 million to 30
million and per capita income of Rs 1000 to Rs.2500, the regression model shows the
relationships that existed between these extremes in the two independent variables. If
the sales forecast is to be made on the basis of values of independent variables falling
outside the above ranges, then the relationships expressed by the regression model may
not hold good. Second, sometimes there may be a lagged relationship between the
dependent and independent variables. In such cases, the values of dependent variables
are to be related to those of independent variables for the preceding month or year as
the case may be. The search for factors with a lead-lag relationship to the sales of a
particular product is rather difficult. One should tryout several indicators before
selecting the one which is most satisfactory. Third, it may happen that the data required
to establish the ideal relationship, do not exist or are inaccessible or, if available, are not
useful. Therefore, the researcher has to be careful in using the data. He should be quite
familiar with the varied sources and types of data that can be used in forecasting. He
should also know about their strengths and limitations. Finally, regression model
reflects the association among variables. The causal interpretation is done by the
researcher on the basis of his understanding of such an association. As such, he should
be extremely careful in choosing the variables so that a real causative relationship can be
established among the variables chosen.
Input-Output Analysis
Another method that is widely used for forecasting is the input-output analysis. Here,
the researcher takes into consideration a large number of factors, which affect the
outputs he is trying to forecast. For this purpose, an input-output table is prepared,
where the inputs are shown horizontally as the column headings and the outputs
vertically as the stubs. It may be mentioned that by themselves input-output flows are of
little direct use to the researcher. It is the application of an assumption as to how the
output of an industry is related to its use of various inputs that makes an input-output
analysis a good method of forecasting. The assumption states that as the level of an
industry's output changes, the use of inputs will change proportionately, implying that
there is no substitution in production among the various inputs. This may or may not
hold good.
The use of input-output analysis in sales forecasting is appropriate for products sold to
governmental, institutional and industrial markets as they have distinct patterns of
usage. It is seldom used for consumer products and services. It would be most
appropriate when the levels and kinds of inputs required achieving certain levels of
outputs need to be known.
A major constraint in the use of this method is that it needs extensive data for a large
number of items which may not be easily available. Large business organisations may be
in a position to collect such data on a continuing basis so that they can use input-output
analysis for forecasting. However, this is not possible m case of small industrial
organisations on account of excessive costs involved in the collection of comprehensive
data. It is for this reason that input, output analysis is less widely used than most
analysts initially expected. A detailed discussion of input-output analysis is beyond the
scope of this book.
Econometric Model
Econometrics is concerned with the use of statistical and mathematical techniques to
verify hypotheses emerging in economic theory. "An econometric model incorporates
functional relationships between estimated techniques into an internally consistent and
logically self-contained framework". Econometrics use both exogenous and endogenous
variables. Exogenous variables are used as inputs into, but they themselves are
determined outside the model. These variables include policy variables controlled
events. In contrast, endogenous variables are those which are determined within the
system.
The use of econometric models is generally found at the macro level such as forecasting
national e and its components. Such models show how the economy or any of its specific
segments operates. As compared to an ordinary regression equation they bring out the
causalities involved more distinctly. This merit of econometric models enables them to
predict turning points more accurately. However, their use at the micro-level for
forecasting has so far been extremely limited.
Applications of research methodology in human resources management
Research methodology widely used in the domains of Human resources Management. It
is called as Human resources Metrics (HR Metrics).
To move to the center of the organization, HR must be able to talk in quantitative,
objective terms. Organizations are managed by data. Unquestionably, at times,
managers make decisions based on emotions as facts. Nevertheless, day-to-day
operations are discussed, planned and evaluated in hard data terms.
Perhaps the most crucial advantage of a sound HR metrics programme is that it enables
HR to converse with senior management in the language of business. Operational
decisions taken by HR are then based on cold, hard facts rather than gut feeling, the
figures being used to back up business cases and requests for resource. The HR function
is transformed from a bastion of 'soft' intangibles into something more 'scientific', better
able to punch its weight in the organisation. In addition, the value added by HR
becomes more visible. This will become increasingly important as more and more
functions attempt to justify their status as strategic business partners rather than merely
cost centres.
The five key practices of the Human Capital Index are as follows:
1. Recruiting excellence
2. Clear rewards and accountability,
3. Prudent use of resources,
4. Communications integrity
5. Collegial flexible workplace
They require the capture of metrics for their very definition. Metrics help
quantify and demonstrate the value of HR
Metrics can help guide workforce strategies and maximize return on HR
investments
Metrics provide measurement standards
Metrics help show what HR contributes to overall business results
SUMMARY
The above lesson has given brief introduction to various applications of various research
techniques in management. It has identified the appropriate tools for applications of
various domains of Management.
KEY TERMS
Marketing research
Brand positioning
Image profile analysis
Market Potential
Demand Measurement
Delphi Method
Time Series analysis
Moving average
HR Metrics
IMPORTANT QUESTIONS
1.
2.
3.
4.
5.
6.
7.
8.
What are all the various domains in which research tools can be used?
Explain the application of Image profile analysis with example.
Differentiate between Primary and secondary research.
What are the limitations of marketing research?
Describe the method for finding out the Market potential.
Explain the Various methods to estimate the Demand.
What do you mean by HR Metrics?
Note down the five key practices of the Human Capital Index.
OBJECTIVES
o
o
o
o
o
o
o
STRUCTURE
Research report
Report format
Data presentation
Pareto chart
RESEARCH REPORT
The final step in the research process is the preparation and presentation of the research
report. A research report can be defined as the presentation of the research findings
directed to a specific audience to accomplish a specific purpose.
Importance of report
The research report is important for the following reasons:
1. The results of research can be effectively communicated to management.
2. The report is the only aspect of the study, which executives are exposed to and their
consecutive evaluation of the project rests with the effectiveness of the written and oral
presentation.
3. The report presentations are typically the responsibility of the project worthiness. So,
the communication effectiveness and usefulness of the information provided plays a
crucial role in determining whether that project will be continued in future.
Steps in report preparation
Preparing a research report involves three steps
1. Understanding the research
2. Organizing the information
3. Writing with effectiveness
Guidelines
The general guidelines that should be followed for any report are as follows:
1. Consider the audience: The information resulting from research is ultimately
important to the management, who will use the results to make decisions. Decision
makers are interested in a clear, concise, accurate and interesting report, which directly
focuses on their information needs with a minimum of technological jargons. Thus, the
report has to be understood by them; the report should not be too technical and not too
much jargon should be used. This is a particular difficulty when reporting the results of
statistical analysis where there is a high probability that few of the target audience have
a grasp of statistical concepts. Hence, for example, there is a need to translate such
terms as standard deviation, significance level, confidence interval etc. into everyday
language.
2. Be concise, but precise: The real skill of the researcher is tested in fulfilling this
requirement. The report must be concise and must focus on the crucial elements of the
project. It should not include the unimportant issues. Researcher should know how
much emphasis has to be given to each area.
As a rough guide, any report of several sections that totals more than 6/to 10 pages
should have a table of contents. The table of contents lists the essence of topics covered
in the report, along with page references. Its purpose is to aid readers in finding a
particular section in the report. If there are many tables, charts, or other exhibits, they
should also be listed after the table of contents in a separate table of illustrations.
3. Executive summary
An executive summary can serve two purposes. It may be a report in miniature covering
all the aspects in the body of the report, but in abbreviated form, or it may be a concise
summary of major findings and conclusions including recommendations.
Two pages are generally sufficient for executive summaries. Write this section after the
report is finished. It must exclude the new information but may require graphics to
present a particular conclusion.
Expect the summary to contain a high density of significant terms since it is repeating
the highlights of report. A good summary should help the decision make and it is
designed to be action oriented.
4. Introduction
The introduction prepares the reader for the report by describing the parts of the
project: the problem statement, research objectives and background material.
The introduction must clearly explain the nature of decision problem. It should review
the previous research done on the problem.
5. Problem statement
The problem statement contains the need for the research project. The problem is
usually represented by a management questions. It is followed by a more detailed set of
objectives.
6. Research objectives
The research objectives address the purpose of the project. These may be research
question(s) and associated investigative questions.
7. Background
The Background material may be of two types. It may be the preliminary results of
exploration from an experience survey, focus group, or another source. Alternatively it
could be secondary data from the literature review. Background material may be placed
before the problem statement or after the research objectives. It contains information
pertinent to the management problem or the situation that led to the study.
8. Methodology
The purpose of the methodology section is to describe the nature of the research design,
the sampling plain, data collection and analysis procedure. Enough details must be
conveyed so that the reader can appreciate the nature the methodology used, yet the
presentation must not be boring overpowering. The use of technical jargon must be
avoided.
9. Research design
The coverage of the design must be adapted to the purpose. The type of research
adapted and reason for adapting that particular type should d be explained.
10. Sampling design
The research explicitly defines the target population being studied and the sampling
methods used. It has to explain the sampling frame, sampling method adapted and
sample size. Explanation of the sampling method, uniqueness of the chosen parameters
or other relevant points that need explanation should be covered with brevity.
Calculation of sample size can be placed either in this part or can be placed in an
appendix.
11. Data collection
This part of report describes the specifics of gathering the data. Its content depends on
the selected design. The data collection instruments (questionnaire or interview
schedule) field instructions can be placed in the appendix.
12. Data analysis
This section summarizes the methods used to analyze the data. Describe data handling,
preliminary analysis, statistical tests, computer programs and other technical
information. The rationale for the choice of analysis approaches should be clear. A brief
commentary on assumptions and appropriateness of use should be presented.
13. Limitations
Every project has weakness, which need to be communicated in a clear and concise
manner. In this process, the researcher should avoid belaboring minor study weakness.
The purpose of this section is not to disparage the quality of the research project, but
rather to enable the reader to judge the validity of the study results.
Generally the limitations will occur in sampling, no response inadequacies and
methodological weakness. It is the researcher's professional responsibility to clearly
inform the reader of these limitations.
14. Findings
The objective of this part is to explain the data rather than draw conclusions. When
quantitative data can be presented, this should be done as simply as possible with
charts, graphics and tables.
The findings can be presented in a small table or chart on the same page. While this
arrangement adds to the bulk of the report, it is convenient for the reader.
15. Conclusions
It can be further divide into two parts as summary and recommendations.
16. Summary and conclusions
The summary is brief statement of the essential findings. The conclusion should clearly
link the research findings with the information needs, and based on this linkage
recommendation for action can be formulated. In some research works the conclusions
were presented in a tabular form for easy reading and reference. The research questions
/objectives will be answered sharply in this part.
17. Recommendations
The researcher's recommendations may be weighed more heavily in favor of the
research findings. There are few ideas about corrective actions. The recommendations
are given for managerial actions rather than research action. The researcher may offer
several alternatives with justifications.
18. Appendices
The purpose of appendix is to provide a place for material, which is not absolutely
essential to the body of the report. The material is typically more specialized and
complex than material presented in the main report, and it is designed to serve the
needs of the technically oriented reader. The appendix will frequently contain copies of
the data collection forms, details of the sampling plan, estimates of statistical error,
interviewer instructions and detailed statistical tables associated with the data analysis
process. The reader who wishes to learn the technical aspects of the study and to look at
statistical breakdowns will want a complete appendix.
19. Bibliography
The use of secondary data requires a Bibliography. Proper citation, style and formats are
unique to the purpose of the report. The instructor, program, institution, or client often
specifies style requirements. It will be given as footnote or endnote format. The author
name, title, publication, year, pager number are the important elements of bibliography.
DATA PRESENTATION
The research data can be presented in Tabular & Graphic form.
Tables
The tabular form consists of the numerical presentation of the data. Tables should
contain the following elements:
1. Table number, this permits easy location in the report
2. Title: the title should clearly indicate the contents of the table or figure.
3. Box head and sub head: the box head contains the captions or labels to the column in
a table, while the sub head contains the labels for the rows.
4. Footnote: footnote explains the particular section or item in the table or figure.
Graphics
The graphical form involves the presentation of data in terms of visually interpreted
sizes. Graphs should contain the following elements:
1. Graph or figure number
2. Title
3. Footnote
4. Sub heads in the axis
Bar chart
A bar chart depicts magnitudes of the data by the length of various bars which have been
laid out with reference to a horizontal or vertical scale. The bar chart is easy to construct
and can be readily interpreted.
Column chart
These graphs compare the sizes and amounts of categories usually for the same time.
Mostly places the categories on X-axis and values on Y-axis.
Pie chart
The pie chart is a circle divided into sections such that the size of each section
corresponds to a portion of the total. It shows the relationship of parts to the whole.
Wedges are row values of data. It is one form of area chart. This type is often used with
business data.
Line Graph
Line graphs are used chiefly for time series and frequency distribution. There are several
guidelines for designing the line graph.
I. Radar diagram
In this the radiating lines are categories; values are distances from the center. It can be
applied where multiple variables used.
Purpose of a Histogram
A histogram is used to graphically summarize and display the distribution of a process
data set. The purpose of a histogram is to graphically summarize the distribution of a
uni-variate data set.
The histogram graphically shows the following:
1. center (i.e. the location) of the data;
2. spread (i.e. the scale) of the data;
3. skewness of the data;
4. presence of outliers; and
5. presence of multiple modes in the data.
These features provide strong indications of the proper distributional model for the
data. The probability plot or a goodness-of-fit test can be used to verify the
distributional model.
Sample Bar Chart Depiction
PARETO CHART
Purpose of a Pareto Chart
A Pareto Chart is used to graphically summarize and display the relative importance of
the differences between groups of data. A bar graph used to arrange information in such
a way that priorities for process improvement can be established.
Sample Pareto Chart Depiction
Purposes
- To display the relative importance of data.
- To direct efforts to the biggest improvement opportunity by highlighting the vital few
in contrast to the useful many.
Pareto diagrams are named after Vilfredo Pareto, an Italian sociologist and economist,
who invented this method of information presentation toward the end of the 19th
century. The chart is similar to the histogram or bar chart, except that the bars are
arranged in decreasing order from left to right along the abscissa. The fundamental idea
behind the use of Pareto diagrams for quality improvement is that the first few (as
presented on the diagram) contributing causes to a problem usually account for the
majority of the result. Thus, targeting these "major causes" for elimination results in the
most cost-effective improvement scheme.
How to Construct
Determine the categories and the units for comparison of the data, such as
frequency, cost, or time.
Total the raw data in each category, then determine the grand total by adding the
totals of each category. Re-order the categories from largest to smallest.
Determine the cumulative percent of each category (i.e., the sum of each category
plus all categories that precede it in the rank order, divided by the grand total and
multiplied by 100).
Draw and label the left-hand vertical axis with the unit of comparison, such as
frequency, cost or time.
Draw and label the horizontal axis with the categories. List from left to right in
rank order.
Draw and label the right-hand vertical axis from 0 to 100 percent. The 100
percent should line up with the grand total on the left-hand vertical axis.
Beginning with the largest category, draw in bars for each category representing
the total for that category.
Draw a line graph beginning at the right-hand corner of the first bar to represent
the cumulative percent for each category as measured on the right-hand axis.
Analyze the chart. Usually the top 20% of the categories will comprise roughly
80% of the cumulative total.
histograms can now be generated on computers has meant that Stemplots are less used
today than in the 1980's, when they first became widely used.
To construct a Stemplot, the observations must first be sorted in ascending order. Here
is the sorted set of data values that will be used in the example:
54 56 57 59 63 64 66 68 68 72 72 75 76 81 84 88 106
Next, it must be determined what the stems will represent and what the leaves will
represent. Typically, the leaf contains the last digit of the number and the stem contains
all of the other digits. In the case of very large or very small numbers, the data values
may be rounded to a particular place value (such as the hundreds place) that will be
used for the leaves. The remaining digits to the left of the rounded place value are used
as the stems.
In this example, the leaf represents the ones place and the stem will represent the rest of
the number (tens place and higher).
The Stemplot is drawn with two columns separated by a vertical line. The stems are
listed to the left of the vertical line. It is important that each stem is listed only once and
that no numbers are skipped, even if it means that some stems have no leaves. The
leaves are listed in increasing order in a row to the right of each stem.
next 6-7, and the last leave 8-9. Here is an example of a split stem plot (using the same
data set from the example above) in which each stem is split into two:
SUMMARY
A research report can be defined as the presentation of the research findings directed to
a specific audience to accomplish a specific purpose.
General guidelines followed to write the report are
1.
2.
3.
4.
The main elements of report are title page, table of contents, executive summary,
introduction, methodology, findings, conclusions, appendices, and bibliography.
Tables and graphs are used for the presentation of data. Different types of graphs are
available like bar chart, pie chart, line chart, area diagram, radar diagram, scatter
diagram, etc. According to the nature of data and requirement the type of graph can be
selected and used effectively.
KEY TERMS
Executive summary
Sampling
Bibliography
Appendix
Interview schedule
Area chart
Line graph
Bar chart
Pie chart
Scatter diagram
Radar diagram
Surface diagram
Pareto Chart
Pictograph
Stem-graph
IMPORTANT QUESTIONS
1.
2.
3.
4.
5.
6.
7.
8.
9.
REFERENCE BOOKS
1. Ramanuj Majumdar, Marketing research, Wiley Estern Ltd., New Delhi, 1991.
2. Harper W Boyd, Jr etal, Marketing Research, Richard D. Irevin Inc. USA 1990.
3. Paul E. Green et al, Research for Marketing Decisions, Prentice Hall of India Pvt.
Ltd., New Delhi, 2004.
- End of Chapter -