Copy Right : Ra i Unive rsit y

11.556 147
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
So far we have discussed a variety of tests that make inferences
about a population parameter such as the mean or the population
proportion. These are termed parametric tests and use parametric
statistics from samples that come from the population being
tested. To use these tests we make several restrictive assumptions
about the populations from which we drew our samples. For
example we assumed that our underlying population is normal.
However underlying populations are not always normal and in
these situations we need to use tests, which are not parametric.
Many such tests have been developed which do not make restrictive
assumptions about the shape of the population distribution.
These are known as non-parametric or distribution free tests.
There are many such tests, we shall learn about some of the more
popular ones.
Non Parametric Tests
What are Non Parametric Tests?
Statistical tests that do not require the estimate of population
variance or mean and do not state hypotheses about parameters
are considered non-parametric tests. When do you use non-
parametric tests?Non-parametric tests are appropriately used when
one or more of the assumptions underlying a particular parametric
test has been violated (generally normality or homoscedasticity).
Generally, however the t-test is fairly robust to all but the severest
deviations from the assumptions.
How do you know if the data are normally distributed? There are
several techniques are generally used to find out if a population
has an underlying normal distribution. these include: goodness
of fit (low power), graphical assessment (not quantitative),
Shapiro-Wilks W (n<50), or DAgostino-Pearson Test K2
(preferred).
As noted non parametric tests do not make the assumption of
normality about the population distribution. The hypotheses of
a non-parametric test are concerned with something other than
the value of a population parameter.
The main advantage of non-parametric methods is that they do
not require that the underlying population have a normal or any
other shaped distribution. The main disadvantages of these tests
are that they ignore a certain amount of information. For example
to convert data to non parametric data we can convert numerical
data to non parametric form by replacing numerical values such as
113.45, 189.42, 76.5, 101.79 by either ascending or descending
order ranks. Therefore we can replace them by 1,2, 3, 4, and 5.
How ever if we represent 189.42 by 5, we lose some information,
which is contained in the value 189.42. 189.42 are the largest value
and this represented by the rank 5. However the rank 5 could also
represent 1189.42, as that would also be the largest value. Therefore
use of ranked data leads to some loss of information.
The second disadvantage is that these tests are not as sharp or
efficient as parametric tests. The estimate of an interval using a
non-parametric test may be twice as large as for the parametric case.
When we use nonparametric tests we trade off sharpness in
estimation with the ability to make do with less information and
to calculate faster.
What happens when we use the wrong test in the wrong situation?
Generally, parametric tests are more powerful than non-parametric
tests (e.g., the non-parametric method will have a greater probability
of committing a Type II error - accepting a false null hypothesis)
Exercise
1. What is the difference between the kinds of questions answered
by parametric tests and those by non-parametric tests?
Important Types of Nonparametric Tests:
Since the theory behind these tests is beyond the scope of our
course we shall look at relevant applications and methodologies
for carrying out some of the more important non parametric
tests.
Non-parametric tests are frequently used to test hypotheses with
dependent samples.
The dependent sample tests are:
1. The signed test of paired data where positive and negative
signs are substituted for quantitative values.
2. Mcnemar test
3. Cochran
4. Wilcoxon test
Non-Parametric Tests for Independent Samples are:
1. Chi square test
2. Kolmogorov Smirnov one sample test
Of these tests we shall cover the McNemar test, the Mann Whitney
U test, the Kologomorov smironov test and the Wilcoxon test.
Table of equivalent parametric and non-parametric tests:
Parametric Non-Parametric
Independent T-Test Mann-Whitney U Test
Paired T-Test Wilcoxon Signed-Rank Test

When do We use Non Parametric Tests?
We use non-parametric tests in least one of the following five
types of situations :
1. The data entering the analysis are enumerative; that is, counted
data represent the number of observations in each category or
cross-category.
2. The data are measured and/ or analyzed using a nominal scale
of measurement.
3. The data are measured and/ or analyzed using an ordinal scale
of measurement.
LESSON 24:
NON PARAMETRIC TESTS
Copy Right : Ra i Unive rsit y
148 11.556
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
4. The inference does not concern a parameter in the population
distribution; for example, the hypothesis that a time-ordered
set of observations exhibits a random pattern.
5. The probability distribution of the statistic upon which the
analysis is based is not dependent upon specific information
or conditions (i.e., assumptions) about the population(s) from
which the sample(s) are drawn, but only upon general
assumptions, such as a continuous and/ or symmetric
population distribution.
According to these criteria, the distinction of non-parametric is
accorded either because of the
1. The level of measurement used or required for the
analysis, as in types 1,2, 3 . That is we use either counted
or ordinal or nominal scale data.
2. The type of inference, as in type 4. We do not make
inferences about population parameters such as the mean.
3. The generality of the assumptions made about the
population distribution, as in type 5. That is we do not
know or make assumptions about the specific form of
the underlying population distribution.
Non-parametric vs. Distribution-free Tests:
As we have seen non-parametric tests are those used when some
specific conditions for ordinary tests are violated? Distribution-
free tests are those for which the procedure is valid for all different
shapes of the population distribution.
For example, the Chi-square test concerning the variance of a
given population is parametric since this test requires that the
population distribution be normal. On the other hand the Chi-
square test of independence does not assume normality condition,
or even that the data are numerical. The Kolmogorov-Smirnov
test is a distribution-free test, which is applicable to comparing
two populations with any distribution of a continuous random
variable.
Standard uses of Non Parametric Tests
Mann-Whitney
To be used with two independent groups (analogous to the
independent groups t-test) we may use the Mann-Whitney Rank
Test as a non-parametric alternative to Students T-test when one
does not have normally distributed data.
Wilcoxon
To be used with two related (i.e., matched or repeated) groups
(analogous to the dependent samples t-test)
Kruskall-Wallis
To be used with two or more independent groups (analogous to
the single-factor between-subjects ANOVA)
Friedman
To be used with two or more related groups (analogous to the
single-factor within-subjects ANOVA)
We now look at a few of the non parametric tests in more details
including their applications.
McNemar Test
This test is used for analyzing research designs of the before and
after format where the data are measured nominally. The samples
therefore become dependent or related samples. The use of this
test is limited to the case where a 2x2 contingency table is involved.
The test is most popularly used to test response to a pre and post
situation of a control group.
We can illustrate its use with an example:
A survey of 260 consumers was taken to test the effectiveness of
mailed coupons and its effect on individuals who changed their
purchase rate for the product. The researcher took a random sample
of consumers before the release of the coupons to assess their
purchase rate. On the basis of their responses they were divided
into groups as to their purchase rate (low, high). After the campaign
they were again asked to complete the forms and again classified
on their purchase rate. Table 1 shows the results from our sample.
Table1
After campign
Before
campign
High purchse rate Low purchase
rate
Ttotal
210


150


Low
purchase
rate
High
purchse
rate

70(A)


80(C)
180(B)


30(D)
260
Cases that showed a change in the before and after the campaign
in terms of their purchase response were placed in cells A and D.
this was done as follows:
An individual is placed in cell A if he or she changed
from a low purchase to a high purchase rate.
Similarly hes placed in D if he goes from high to a low
rate.
If no change is observed in his rate he is placed on cells
BorC.
The researches wishes to determine if the mail order campaign
was a success. We shall now briefly outline the various steps
involved in applying the McNemar test.
Step1
We state the null hypotheses. This essentially states that there is
no perceptible or significant change in purchase behavior of
individuals. Thus for individuals who change their purchase rate
this means that the probability of those changing from high to
low equals low to high. This equal to .5.
Ho: P(A)=P(D)
Ha: P (A)P(D)
To test the null hypotheses we would examine the cases of change
from cells A to D.
Step2
The level of significance is chosen, for example = .05
Copy Right : Ra i Unive rsit y
11.556 149
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
Step3
We now have to decide on the appropriate test to be used. The
McNemar test is appropriate because the study is a before and
after study and the data are nominal. The study involves the study
of two related variables.
The McNemar test involves calculating the chi square value as
given by the formula below:
Chi square=( |A-D | -1)
2
/ A+D
That is we calculate the absolute difference between the A and D
cells.
Step 4
The decision rule
For a=. 05, the critical value of
2
is 3.84 for degree of freedom 1.
Therefore we will reject the null hypotheses if the calculated
2
exceeds the critical value from the tables.
Step 5
We now actually calculate the test statistic:
The calculated ?
2
=( | A-D | -1)
2
/ A+D =(70-30 -1)
2
/ 100=15.21
Step 6
Draw a statistical conclusion.
Since calculated ?
2
exceeds the critical value
,
we reject the null
hypotheses and we can infer that the mail coupon campaign was
successful in increasing the purchase rate of the product under
study.
When an analysis involves more than two variables we use the
Cochran Q test. For situations where involving repeated
observations where the dependent variable can take on only two
values; either 0 or 1.
Tests for Ordinal Data
So far the test we have discussed is applicable only to nominal
data. We now look at a test, which is specifically designed for
ordinal data.
Kolmogorov Smirnov one Sample Test
This test is similar to the chi square test of goodness of fit. This
is because it looks at the degree of agreement between the
distribution of observed values and some specified theoretical
distribution (expected frequencies). The Kolmogorov Smirov test
is used if we want to compare the distribution on an ordinal scale.
We can look at an example to see how this test works:
A paint manufacturer is interested in testing for four different
shades of a colour: very light, very light, bright and dark. Each
respondent is shown four prints of the shades and asked to
indicate his preference. If colour shade is unimportant, the photos
of each shade should be chosen equally. Except for random
differences. If colour shades are important the respondents should
prefer one of the extreme shades.
Since shade represents a natural ordering, the Kolmogorov test is
applied to test the preference hypothesis. The test involves
specifying the cumulative frequency distribution. We then refer to
the sampling distribution indicates where there is a divergence
between the two distributions is likely to occur due to chance or if
the observed difference is due to a result of preference.
Suppose for example that a sample of 200 homeowners and we
got the following shade preference distribution;
Very light: 80
Bright: 40
Dark: 20
The manufacturer asks whether these results indicate a preference.
The data is shown in table 2 below.

Rank of shade chosen
Very light light bright dark
F=no. of
homeowners
choosing that
rank

Fo(X)=
theoretical
cumulative
distribution of
choices under
Ho

Sn(X)=
cumulative
distribution of
observed
choices
| Fo(X)-Sn(X)|


80




.25





.40




.15

60




.50





.70




.20


40




.75





.90




.15

20




1.00





1.00




0.0
We would carry out the test as follows:
1. Specification of null hypotheses; Ho is that there would be no
difference among the shades. Ha: there is a difference in the
shades of the new colour.
2. 2. The level of significance: The test would be conducted at the
55 level.
3. Decision regarding statistical test. Here the Kolmogorov t-
Smirnov test is appropriate because the data measured are ordinal
and we are interested in comparing the above frequency
distribution with a theoretical distribution.
4. This test focuses on the largest value of the deviations among
observed and theoretical proportions.
D= max | Fo (X)-Sn (X)|
Where Fo (X) is the specified cumulative frequency distribution
under Ho for any value of x and is the proportion of cases expected
to have scores equal to or less than x.
So(X): observed cumulative frequency distribution of a random
sample of N observations where X is any possible score.
5. The decision rule
If the researcher chooses a =. 05, the critical value of D for
large samples is given by the formula 1.36n where n is the
sample size. For example the critical value is .96. the decion rule
therefore states that null hypotheses will be rejected if computed
D >.96.
Copy Right : Ra i Unive rsit y
150 11.556
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
6. Calculation of test Statistic
The theoretical probability distribution is calculated by taking
each class frequency as if it were under the null hypotheses, i.e.,
in our case they would be equal. It is then expressed as a fraction
of the total sample. Thus row 2 would be calculated as 50/
200=. 25, 100/ 200=. 5, 150/ 200=. 75, 200/ 200=1
The observed cumulative frequency of row 3 is found by 80/
200=. 40, 140/ 200=. 70, 180/ 200=. 90
The calculated D value is the point of greatest divergence between
cumulative observed and theoretical distributions. In our example
this is .20
7. Drawing a statistical conclusion: Since calculated D value (.20)
exceeds the critical value of .096, the null hypotheses of no
difference among shades are quickly rejected.
Mann Whitney U Test
This is used for data, which is ordinal and can be ranked. This test
makes use of the actual ranks of the observations as a means of
testing the hypotheses about the identity of two population
distributions.
An Example will Illustrate this Test
To illustrate its use we will examine regular and commercial account
satisfaction data from Table 3. The table contains the attitude
scores obtained from 30 customers (15 regular accounts and 15
commercial accounts). The scores from the combined samples
were then ranked in terms of their magnitude of the original
score (columns 3 and 4 of Table3). That is, the highest score gets
rank 1, the next highest 2, and so on.
The Null Hypothesis
Ho is that there is no difference in attitudes of the two groups of
accounts towards bank services.. Ha is that there is a significant
difference difference in attitudes of the two groups.
The level of significance for this test is .05.
Ranking of
raw scores




Regular commercial
Regular
accounts

19
70
77
14
83
87
68
72
76
90
26
66
60
81
46

Commercial
accounts

48
52
69
13
73
15
50
61
21
47
80
36
78
71
65




27
12
7
29
3
2
14
10
8
1
25
15
18
4
23




21
19
13
30
9
28
20
17
26
22
5
24
6
11
16
The Mann whiteny Test is used because the data is ordinal and
converted into ranks.. Also the samples are independent.
The formula for the Mann whitney U value is :
U
1=
n
1
70n2 +n
1
(n
1
+ 1)/ 2 - R
U2 = n1 n2 - U1
Analysis of Differences
Where n1 and n2 are the two sample sizes and R1 and R2are the
sums of the ranks for each group. Letting regular accounts be
sample 1 and the commercial accounts be sample 2, we find that
n1 = 15 n2=15 R1 = 198 R2= 267
The critical value for the statistic U* is found in Appendix I. For
on = 0.05, n1 = 15 and n2 = 15, the critical value for the Mann-
Whitney statistic is U* = 64 for a two-tailed test. For this test, the
null hypothesis will be rejected if the computed value, U, is 64 or
less. Otherwise, it will not be rejected. This decision is just the
opposite of the decision making procedure wefollowed for most
of the other tests of significance.
Calculation of the Test Statistic.
Therefore,
U1 = (15)(15) + 15(15 + 1) - 198
2
= 225 + 120 - 198
= 345 - 198
= 147
And
U2 = (15)(15) - 147 = .225 - 147
= 78
The Mann-Whitney U is the smaller of the two U values calculated.
In this case the Mann-Whitney Uis 78. Inspection of the formulas
for calculating U will indicate that the more similar the two groups
are in their attitudes, the smaller the R values will be and the larger
the value of U. Therefore weare testing the probability of obtaining
a value the size of the smallest of the two groups if the two
groups are indeed similar in their attitudes.
Drawing a Statistical Conclusion:
Since the computed U is larger than the critical U*, it does not fall
in the critical region and the null hypothesis is not rejected. For
this test the computed value must be less than the critical value U*
to reject the null hypothesis. Once again, the evidence does not
support a difference in the attitudes between the two groups.
The Kruskal- Wallis test is an extension of the Mann-Whitney U
t situations where more than two independent samples are being
compared example, this test could be used if the customers were
divided into three or groups based on some criterion, such as
regular accounts, commercial aCC4 and charge accounts.
Signed and Rank or Wilcoxon test
This test is the complement of the Mann Whitney U test and is
used when ordinal data on two samples are involved and the two
samples are related or dependent. This test is therefore suited to a
pre-test and post-test situation. The test can also be used for
interval or ratio data when the assumptions underlying the
parametric z or t test cannot be met.
Copy Right : Ra i Unive rsit y
11.556 151
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
An example illustrates this rule:
To illustrate the procedure, suppose that our bank in the previous
examples wanted to test the effectiveness of an ad campaign
intended to enhance the awareness of the banks service features.
The bank administered a questionnaire before the ad campaign
designed to measure the awareness of services offered. After the
ad campaign, the bank administered the same questionnaire to
the same group of people. The bank wished to determine whether
there was any change in the awareness of services offered due to
the ad campaign. Both the before-and after ad campaign scores is
presented in Table 4
Table 4
Table 4
Consumer Awareness of Bank
Services offered

Before Ad After Ad Difference overall
ConsumerCampaign Campaign In Score
Rank of
Rank of

Rank
+ Scores Scores
1 82 87 5 6.5 6.5
2 81 4 3 4 4
3 89 84 -5 6.5
4 74 76 2 2 2.5 6.5
5 68 78 10 9 9
Solution
Step 1
The Null Hypothesis. The null hypothesis to be tested is that
there is no difference in awareness of services offered after the ad
campaign. The alternative hyp6thesis would be that there was an
awareness of the services after the ad campaign.
Step 2The Level of Significance
It was decided that = 0.05. Step J.
The Statistical Test
The Wilcoxon test is appropriate because the study is of related
samples and in which the data measured, is ordinal and the
differences can be ranked in magnitude. The test statistic calculated
is the T value. Since the direction of the difference is predicted, a
one-tailed test is appropriate.
Step 4. The Decision Rule
The critical value of the Wilcoxon T is found Appendix J for n =
10 at the 0.05 level of significance and a one-toiled test are 10. This
indicates that a computed Tvalue of less than 10, the critical value,
rejects the null hypothesis. The argument is similar to that, of the
Mann-Whitney U statistic.
Step 5. Calculate the Test Statistic
The procedure for the test is very simple. The signed difference
between each pair of observations is found. Then these differences
are ranked-ordered without regard to their algebraic sign. Finally,
the sign of the difference is attached to the rank for that difference.
The test statistic, T, is the smaller of the two sums of the ranks.
For our example, T = 6.5 since the smaller sum is associated with
the negative difference.
If the null hypothesis is true the sums of positive and negative
ranks should be approximately equal. However the larger the
difference between the underlying population, the smaller would
be the value of T since it is defined as the smaller of the ranks.
Step6 We draw a statistical conclusion;
since the computed T value is 6.5 is less than the critical T valie of
120; the null hypothesis, which states that there is no difference in
the awareness of bank services, is rejected.
Exercises
Small Grocers Association
The Small Grocers Association is a group of independent grocers
who have banded together so that they may compete with the
larger supermarket chains. By making large purchases as a group, it
is able to pay less than if it were to purchase as an individual store.
Because of its newfound purchasing power, the association wants
to advertise that its prices are really no high; the neighborhood
stores that is a member of the association than in supermarkets.
It has chosen one Eroducta 5-pound ham-with which to make
its point in the advertisement.
At its recent monthly meeting, the executive board brought the
advertising campaign to the members for approval. One member
stated that we better make sure of our claim.
The board of directors agreed, and asked the associations
marketing director to check this out before the copy was submitted
to the newspaper. To check on this claim, the marketing director
obtained a random sample of prices from six neighborhood
association member stores and nine supermarkets
Discussion Questions
1. What analytical technique would be appropriate for making 0
formed analysis?
2 What ore the null and the alternative hypotheses to be tested?
3. Run an analysis of the data using the 0.05
National Motors, Inc
National Motors is a manufacturer of motor scooters. As part of
their operating policy, the executives wished to determine whether
there was a difference between the dealers customers and the
companys dealers in terms of satisfaction with the companys
warranty policy. National Motors marketing research department
developed a questionnaire utilizing Liket-type statements that
encompassed a full range of service and warranty questions. The
researchers believed the data obtained from the questionnaires
were ordinal. The questionnaires were mailed to a random sample
of customers who had returned the National Motorss warranty
card and a second mailing of the questionnaires were sent to a
sample
Copy Right : Ra i Unive rsit y
152 11.556
R
E
S
E
A
R
C
H

M
E
T
H
O
D
O
L
O
G
Y
Customers Dealers Customers Dealers
74 92 43 49
81 42 23 32
35 54 88 52
59 59 55 27
90 83 67 81
33 30 53 68
82 34 85 51
68 54 70 65
56 39 30 25
46 65 75 46


Discussion Questions
1. What analytical technique(s) would beappropriate for making a
formal analysis?
2. What is the null hypothesis to betested?
3. Run an analysis of the data using the 0. .05 level of significance.
What conclusions can you draw?
Ref erences
1. Levin and Rubin management Statistics
2. Luck and rubin Marketing Research
3. Dr Ashrams web site
Notes