Vous êtes sur la page 1sur 19

Statistics for Marketing & Consumer Research (M.

Mazzocchi)
SAGE Publications

Multiple choice questions

Chapter 1

1. The difference between random and systematic errors is:


A) Systematic errors are always larger than random errors
B) Random errors are subject to probability laws, while systematic errors
are not
C) Systematic errors can be eliminated by taking repeated measurements

2. If measurement error follows a Normal curve centred at zero, then:


A) Errors in excess compensate error in defect
B) Errors in excess are more likely than error in defect
C) The average error is negative

3. Which of the following is a qualitative ordinal variable?


A) Monthly income
B) Quantity of coffee drank in a week
C) Customer satisfaction measured on a Likert scale from one to seven

4. High reliability for a set of questionnaire items measuring a latent


construct means that:
A) The items are bipolar
B) The items are not internally consistent
C) The items are internally consistent

5. What is the difference between a Likert and a Semantic Differential (SD)


scale?
A) The Likert scale requires two bipolar attributes, the SD scale one
B) The SD scale requires two bipolar attributes, the Likert scale one
C) There is no difference

Chapter 2

6. Primary data are:


A) Data already collected by others for different purposes
B) Data collected before secondary data
C) Data explicitly collected for the specific purpose of the research

7. Secondary data are always preferable to primary data when:


A) Secondary data are more expensive than primary data
B) The target population and sampling fit with the research objective and
quality is acceptable
C) The sample size of secondary data is larger than the one of primary data
8. What is the difference between household budget surveys and panel
household surveys?
A) Household budget surveys include the same households over time
B) Panel household surveys include the same households over time
C) Panel household surveys do not record budgets

9. What is the COICOP classification?


A) A classification of households according to the economic status of the
reference person
B) A classification of surveys according to the type of sampling procedure
C) A classification of expenditure items according to their purpose

10. What is scan data?


A) Purchase data collected through a bar-code scanner
B) Old data transcribed into electronic files through scanning facilities
C) Machine-readable questionnaires

Chapter 3

11. Which of the following is not a non-response error?


A) Sampling frame error
B) Not-at-home error
C) Refusal

12. What is the difference between sampling and non-sampling error?


A) Sampling error can be quantified with probabilistic sampling
B) Non-sampling error can be quantified with large sample sizes
C) Sampling error is always larger than non-sampling error

13. What is the reference population?


A) A population used for comparison with the selected sample
B) The list of subjects included in the sample
C) The complete set of subjects relevant to the research

14. Which survey method has the highest non-response rate?


A) Telephone
B) Mall intercept
C) Mail surveys

15. What is the duration of a telephone interview to ensure adequate quality


of responses?
A) 15-20 minutes
B) 30-40 minutes
C) 45-60 minutes

16. What is the duration of a telephone interview to ensure adequate quality


of responses?
A) 15-20 minutes
B) 30-40 minutes
C) 45-60 minutes

17. Sensitive questions


A) Should never be at the beginning of a questionnaire
B) Should never be asked through multiple choice questions
C) Are better recorded through face-to-face interview

Chapter 4

18. What is a dummy variable?


A) A metric variable with no decimals
B) A binary variable
C) A useless variable which should be discarded

19. Which of the following is not the definition of an outlier?


A) A value with a distance from the mean higher than 2.5 times the
standard deviation
B) A value with a distance from the mean higher than 1.5 times the
interquartile range
C) A value with distance from the mean higher than the median

20. Listwise or casewise deletion of missing data implies:


A) That observations (cases) with missing data are omitted from the
analysis
B) That observations (cases) enter the analysis only with their non-missing
values
C) That variables with missing data are omitted from the analysis

21. In statistical analysis with pairwise deletion of missing data:


A) Observations (cases) with one or more missing data are omitted from
the analysis
B) In each estimation step all valid cases are exploited
C) If a variable is omitted in an estimation step because of missing data,
then it does not enter estimation any more

22. Missing response should be treated as non-random when:


A) They are more than 5% of total responses
B) The mean of relevant variables is very different between respondents
and non-respondents
C) The mean of relevant variables for non-respondents is equal to the one
for respondents

23. Which of the following is not a good strategy for imputing missing data?
A) Substituting missing values with the sample mean for that variable
B) Substituting missing values with the standard deviation for that variable
C) Imputing missing values with several approaches, then taking the
average

24. Which of the following is a measure of central tendency?


A) Standard error
B) Coefficient of variation
C) Mode

25. What is the median?


A) The value which splits the observations in a data-set in two halves, those
below and those above the median value
B) The most likely value
C) The average value

Chapter 5

26. What is a statistic sample?


A) A measure of variability
B) A group of variables measured on all the subjects object of interest
C) A subset of the target population expected to represent the whole
population

27. What is a sampling frame?


A) A list of all subjects included in the sample
B) A list of statistics to be computed in the sample
C) A list of all subjects included in the reference population

28. Which of the following is a form of probability sampling?


A) Convenience sampling
B) Quota sampling
C) Stratified sampling

29. Which of the following statistics is a measure of sample precision?


A) Standard error of the mean
B) Mode
C) Mean

30. Which of the following parameters influence the choice of the sample
size?
A) Variability of the target variable in the population
B) Size of the population
C) Both

31. What is the advantage of the sampling error over the non-sampling
error?
A) Sampling error can be quantified according to probability laws
B) Sampling error is much smaller than non-sampling error
C) Sampling error can be eliminated by training interviewers

32. The rationale behind stratified sampling is:


A) to maximise heterogeneity within each stratum
B) to minimise heterogeneity between different strata
C) to maximise homogeneity within each stratum

33. A population parameter is estimated in a sample through:.


A) A sample statistic
B) The precision level
C) The sampling frame

34. Which of the following statistics is an estimate of population variability


A) Sample average
B) Population average
C) Sample standard deviation

35. Which of the following situations does not favor the use of a census?
A) There is high variance in the characteristic to be measured.
B) The cost of nonsampling errors is low.
C) The population is large.

36. All of the following statements are limitations of simple random


sampling except:
A) It is often difficult to construct a sampling frame that will permit a
simple random sample to be drawn.
B) Simple random sampling often results in lower precision with larger
standard errors than other probability sampling techniques.
C) The sample results may be projected to the target population.

37. Which probability sampling technique selects a random starting point


and picks up every ith element in succession from the sampling frame.
A) Simple random sampling
B) Systematic sampling
C) Cluster sampling

38. In which one of the following ways do cluster sampling and stratified
sampling differ:
A) the former is probabilistic and the latter is non probabilistic
B) there is no difference
C) with respect to homogeneity and heterogeneity within/across subgroups

39. All of the factors listed below favor the use of probability sampling
except:
A) Nonsampling errors are likely to be an important factor
B) The nature of the research is conclusive
C) The population is heterogeneous with respect to variables of interest

40. The difference between the mean value for the sample and the true
mean value of the population is a measure of :
A) Precision
B) Accuracy
C) Randomness

41. The probability sampling technique where each element in the


population has exactly the same probability of extraction is:
A) Stratified sampling
B) Simple random sampling
C) Quota sampling
42. The sampling technique which divides the population into sub-
populations which are expected to be similar among them is:
A) Stratified sampling
B) Simple random sampling
C) Cluster sampling

43. Which of the following is an estimate of the variability of estimates of


the mean in different samples?
A) Standard error of the mean
B) Variance
C) Standard deviation

Chapter 6

44. What is the probability distribution for sample means extracted through
simple random sampling?
A) The uniform distribution
B) The Normal distribution
C) The F distribution

45. The 95% confidence interval for a mean from a large sample has
A) A width of about twice the standard error of the mean
B) A width of about four times the standard error of the mean
C) A width of about ten times the standard error of the mean

46. The critical values for hypothesis testing are:


A) A measure of variability of the target variable
B) The values which separate the acceptance region from the rejection
region
C) The area under the probability distribution

47. What is the level of significance of a test?


A) The probability of rejecting the null hypothesis when it is actually true
B) The probability of non-rejecting the null hypothesis when it is actually
true
C) The probability of non-rejecting the null hypothesis when it is actually
false

48. What is the relation between level of confidence and the level of
significance ()?
A) The level of confidence is 1+
B) The level of confidence is 1
C) There is no difference

49. What is the power of a test?


A) The probability of rejecting the null hypothesis when it is actually true
B) The probability of non-rejecting the null hypothesis when it is actually
true
C) The probability of rejecting the null hypothesis when it is actually false

50. What is the difference between parametric and non-parametric tests?


A) Parametric tests do not require assumption on the probability
distribution of the variables being tested
B) Non-parametric tests do not require assumption on the probability
distribution of the variables being tested
C) Parametric tests are more powerful

51. If the same sample of individuals is interviewed before and after an


advertising campaign, a mean comparison test about the purchasing habits
before and after the campaign is a test for:
A) Independent samples
B) Related samples
C) Paired samples

52. Which test is more appropriate for mean comparison with related
samples?
A) The t-test
B) The F-test
C) The Wilcoxon test

53. What is the distribution of the ratio of two variances under the null
hypothesis of variance equality?
A) The t-distribution
B) The Normal distribution
C) The F-distribution

54. A t-test in two means returns a t-statistic with a p-value of 0.03. Which
of the following is correct?
A) The means of the two populations are equal at 97% confidence level
B) The means of the two populations are different between each other at
95% confidence level but not at a 99% confidence level
C) The mean of the two populations show a 3% difference

Chapter 7

55. Which of the following is a consequence of running multiple mean


comparison tests?
A) It is more likely to reject the null hypothesis when it is true
B) It is impossible to compute the F-statistic
C) It is less likely to reject the null hypothesis when it is true

56. When does one-way ANOVA rejects the hypothesis of mean equality?
A) When all means are different
B) When at least two means are different
C) When all means are equal

57. What is the main difference between planned comparisons and post-hoc
tests?
A) Planned comparison take into account the familywise error, post-hoc
tests dont
B) Post-hoc tests take into account the familywise error, planned
comparison dont
C) Planned comparison are decided prior to the analysis, post-hoc test
afterwards

58. What is three-way ANOVA?


A) An ANOVA with three variables and one factor
B) An ANOVA with one variable and three factors
C) An ANOVA with both categorical and scale variables

59. What is the effect size in ANOVA?


A) It is the number of factors being considered
B) It corresponds to the F-value
C) Is a measure of the relative weight of the variability imputable to the
factor

60. What is the difference between random and fixed effects?


A) Fixed effects are measured with no error, random effects are the
outcome of a random variable
B) Fixed effects are larger than random effects
C) Random effects are under the control of researchers, fixed effects are
not

61. Which of the following is not an assumptions of one-way ANOVA?


A) Independence between the units sampled in different treatments
B) Absence of large discrepancy in variance across different treatments
C) Different treatments are measured on the same units

62. What is the difference between multi-way (factorial) ANOVA and


multivariate ANOVA?
A) Multivariate ANOVA has more than one target variable, factorial ANOVA
only one
B) Factorial ANOVA has more than one target variable, multivariate ANOVA
only one
C) There is no difference

63. What is the General Linear Model (GLM)?


A) A regression model with the target variable(s) on the left-hand side and
the factors on the right-hand side
B) A general model which includes ANOVA, factorial ANOVA, MANOVA and
ANCOVA as special cases
C) Both of the above

64. Which of the following statement on ANCOVA is true?


A) ANCOVA is a GLM where all the right-hand side variables are
categorical
B) ANCOVA is a GLM where the dependent variable is categorical
C) ANCOVA is a model which can contain both metric and non-metric
variables on the right-hand side
Chapter 8

65. If the F-test in a multiple regression equations returns a probability of


0.22, it means that:
A) None of the coefficient is significantly different from 0
B) All of the coefficients are significantly different from 0
C) At least one coefficient is significantly different from 0

66. The t-test on regression coefficients tests the null hypothesis that:
A) There is no collinearity
B) The residuals are normally distributed
C) The coefficient is significantly different from 0

67. Which of the following statements on the relationship between the


bivariate correlation coefficient r and bivariate regression is true?
A) The regression coefficient is equal to the bivariate correlation coefficient
B) The regression R2 is the squared correlation coefficient
C) The F-test on the regression is equal to the bivariate correlation
coefficient

68. What is the difference between covariance and correlation?


A) Covariance depends on the measurement unit, correlation is
standardized to be between zero and one
B) Correlation depends on the measurement unit, covariance is
standardized to be between zero and one
C) Covariance depends on the measurement unit, correlation is
standardized to be between minus one and one

69. A bivariate correlation coefficient equal to minus one means that:


A) There is no correlation between the two variables
B) The two variables are perfectly correlated, as one increases by n%, the
other also increases by n%
C) The two variables are perfectly correlated, as one increases by n%, the
other decreases by n%

70. Which of the following is an appropriate equation for multiple


regression?
A) yi = a + bxi + ei
B) yi = bx1i + bx2i
C) y = X +

71. The R-square indicator measures the goodness of fit of a regression


model. What does an R-square value of zero mean?
A) There is no relationship between the dependent and explanatory
variables
B) The model performance is very poor
C) The model fits perfectly the data
72. What does the partial correlation coefficient measure?
A) The correlation between two variables after controlling for the effects of
one or more additional variables
B) The correlation among three variables
C) The percentage of correlation between two variables which depends
upon a third variable

73. What is the difference between stepwise and forward model selection
methods?
A) The stepwise method allows the removal of variables after they have
been included, the forward method does not
B) The stepwise method goes backward (removing variables), the forward
method goes forward (adding variables)
C) There is no difference

Chapter 9

74. When are two categorical variables said to be associated?


A) When frequencies in the contingency table are equally distributed
across cells
B) When a relationship exists between frequencies in each category of the
first variable and frequencies in categories of the second variable
C) When they have the same number of categories

75. What is the null hypothesis of the Pearson chi-square test?


A) That two categorical variables are independent
B) That two categorical variables are associated
C) That two categorical variables have the same distribution

76. Which of the following test works with strictly nominal variable?
A) Goodman and Kruskal's Lambda
B) Somers d statistic
C) Tau statistic

77. What is the dependent variable in a log-linear model?


A) The exponent of the cell frequencies in a contingency table
B) The logarithm of the cell frequencies in a contingency table
C) The ratio of the cell frequencies in a contingency table

78. What is the relation between the Pearson chi-square test and log-linear
analysis in a 22 table?
A) The Pearson chi-square test is more powerful
B) Log-linear analysis does not work with 22 tables
C) The Pearson chi-square test corresponds to the test on 2nd order
interaction in log-linear anaysis

79. What is the model selection process in hierarchical log-linear analysis?


A) Higher order interaction are tested first, the process stops when
deletion of a term makes the frequencies predicted by the model
significantly different from frequencies of the saturated model
B) Lower order interaction are tested first, the process stops when deletion
of a term makes the frequencies predicted by the model significantly
different from frequencies of the saturated model
C) Higher order interaction are tested first, the process stops if the
frequencies predicted by the model are not significantly different from
frequencies of the saturated model

80. Which of the following is used in log-linear analysis as a distributional


assumption for the frequencies of the contingency table?
A) Normal distribution
B) Student-t distribution
C) Poisson distribution

81. Which elements are related in canonical correlation analysis?


A) Two categorical variable
B) Two sets of variables
C) Two frequency tables

82. What are canonical cross-loadings?


A) The correlations between a variable in one set and the opposite
canonical variate
B) The correlations between two canonical variates
C) The correlations between a variable in one set and the corresponding
canonical variate

83. What is redundancy analysis?


A) An evaluation of which variables are redundant to canonical correlation
analysis
B) A test of significance on the canonical correlations
C) An evaluation of the proportion of variance explained by canonical
variates

Chapter 10

84. In which of the following aspects is there no difference between factor


analysis and PCA (Principle Component Analysis)?
A) The way scores are determined
B) The indeterminacy of solutions
C) The use of rotation methods

85. Which of the following is an implication of running PCA or factor


analysis on the correlation rather than covariance matrix?
A) It eliminates the influence of different measurement units
B) It increases the number of factors (components) extracted
C) All communalities are equal to 1

86. Which of the following is the Kaiser rule for deciding the number of
components?
A) Extracted components should account for at least 70% of the variability
B) The eigenvalues of the extracted component should be larger than the
mean eigenvalue
C) The scree diagram shows an elbow

87. What does a communality measure?


A) The proportion of original variability of the entire data-set retained by
the extracted factors (components)
B) The proportion of original variability for each variable retained by the
extracted factors (components)
C) The proportion of original variability discarded in the analysis

88. What does a VARIMAX rotation do?


A) It provides an orthogonal rotation of the original solution which
maximizes the variance of loadings in each factor
B) It provides an orthogonal rotation of the original solution which
maximizes the variance of loadings for each variable
C) An average of the two rotations above

89. Which of the following is a definition of factor loading?


A) The correlation between the factor scores and the original variables
B) The correlation between two factors
C) The ratio between two communalities

Chapter 11

90. Which of the following is a requirement for discriminant analysis as a


classification technique?
A) There must be at least three groups
B) The number of variables must be larger than the number of groups
C) The classification of units into groups must be known

91. What is the discriminant function?


A) A function of the classification variable which tests the relevance of
predictors
B) A function of predictors which allows classification into groups
C) A unique value which discriminates between two groups

92. What is the cut-off point in binary discriminant analysis?


A) A unique value which defines classification for each unit according to
the corresponding value of the discriminant function
B) The stopping rule in stepwise discriminant analysis
C) A test on the significance of the predictors

93. To which purpose is the Wilks Lambda used in discriminant analysis?


A) To evaluate the discriminant power of each predictor
B) To test for significantly different means of a discriminant function across
groups
C) Both of the above

94. What is the purpose of the Boxs M statistic?


A) To evaluate the rate of correct predictions
B) To test for significant differences in the discriminant functions across
groups
C) To test the null hypothesis of equality of covariance matrices across
groups

95. How should be the covariance matrices of predictors be across the


groups?
A) Linearly related
B) Significantly different
C) Equal

96. What is cross-validation?


A) A method to compute correlations between groups
B) A method to evaluate the predictive power of the discriminant functions
C) A method to evaluate the discriminating power of the predictors

Chapter 12

97. The objective of cluster analysis when extracting cluster is:


A) Maximising homogeneity within clusters and heterogeneity between
them
B) Maximising heterogeneity within clusters and homogeneity between
them
C) Maximising homogeneity within and between clusters

98. Given N statistical units, an agglomerative hierarchical method


A) Starts with N clusters and proceeds with merges until one cluster is
obtained
B) Starts with one cluster and proceeds with extractions until N clusters
are obtained
C) Extracts immediately the desired number of clusters

99. The Ward algorithm is


A) A non hierarchical method
B) A hierarchical agglomerative method
C) A hierarchical divisive method

100. K-means clustering is a method of:


A) Non-hierarchical clustering
B) Divisive clustering
C) Agglomerative clustering

101. The CCC, Pseudo F and Pseudo t2 are statistics for:


A) Identifying the optimal number of cluster
B) Determining a stopping rule in hierarchical algorithms
C) Testing whether mean differ across clusters

102. Which of the following is a characteristic of the k-mean clustering:


A) It provides output for a range of numbers of clusters
B) It allows reallocation of units at each iteration
C) It allocates the same observation to more than one cluster

103. What is the difference between the centroid and average linkage
methods?
A) Centroid computes all potential distances between two clusters, then
takes the average, while the average linkage computes the mean values
within each cluster, then computes the distance between cluster means
B) Average linkage computes all potential distances between two clusters,
then takes the average, while the centroid method computes the mean
values within each cluster, then computes the distance between cluster
means
C) There is no difference

104. What is the difference between single linkage method and the complete
linkage method?
A) The single linkage considers the minimum distance between
observations belonging to different clusters, the complete linkage the
maximum distance
B) The complete linkage considers the minimum distance between
observations belonging to different clusters, the single linkage the
maximum distance
C) The single linkage considers one distance between two clusters only
(taken randomly), the complete distance considers all distances between
two clusters

Chapter 13

105. What is a proximity matrix


A) A symmetric matrix containing measure of similarity or dissimilarity
between objects
B) A symmetric matrix containing two-dimensional co-ordinates of objects
C) A correlation matrix

106. Which of the following is not a characteristic of non-metric scaling?


A) Deals with ordinal data and guarantees monotonicity
B) Does not work with metric data
C) Has no analytical solution and works through computational algorithms

107. What is the difference between the compositional and decompositional


approach
A) The compositional approach uses ordinal variables, the decompositional
approach uses metric variables
B) The compositional approach requires evaluations of the product
attributes, the decompositional approach requires evaluation of the
product in its integrity
C) Both of the above

108. What are the ideal points?


A) The co-ordinates of the best-selling products
B) The average product evaluations for each subject
C) The points in the perceptual mapping representing the ideal product for
each subject

109. What is the difference between internal and external preference


mapping?
A) Internal preference mapping always precedes external preference
mapping
B) Internal preference mapping bases attribute evaluations on ordinal
preference variables, external preference mapping on metric perceptual
or objective variables
C) Internal preference mapping is decompositional, external preference
mapping is compositional

110. How are perceptions measured?


A) As the ordered ranking of different objects as stated by each subject
B) As the evaluation of different attributes of the objects as stated by each
subject
C) As the paired comparison between pair of objects as stated by each
subject

111. How are preferences measured?


A) As the ordered ranking of different objects as stated by each subject
B) As the evaluation of different attributes of the objects as stated by each
subject
C) As the objective measurement of product attributes

112. Which value of the STRESS function indicate a good fit?


A) A value close to one
B) A negative value
C) Below 0.05

Chapter 14

113. How are distances between categorical variables defined?


A) By considering the nominal variables as scale variables
B) By computing the distance between the actual frequencies and the
expected frequencies assuming no association (like in chi-square
statistics)
C) By considering the differences between the cell frequency and the total
frequency

114. What is total inertia?


A) A measure of association between the two variable in correspondence
analysis
B) The chi-square statistic divided by the number of observations
C) Both of the above

115. What does the percentage of inertia for a given dimension measure?
A) The percentage of total variability explained by that dimension
B) The percentage of the existing association explained by that dimension
C) The percentage of variability not explained by that dimension

116. What is a row mass?


A) The row totals of the contingency table
B) The inertia divided by the number of observations
C) The chi-square statistic for each category

117. What are the implications of a row principal normalization for extracting
the dimensions?
A) That distances between categories of the row variable are assumed to
be more meaningful
B) That distances between categories of the row and column variables are
ignored
C) That distances between categories of the column variables are assumed
to be more meaningful

118. What is the bi-plot?


A) A plot showing all the categories of several categorical variables in the
same space
B) A double graph, one showing the categories of the row variable, one
those of the column variable
C) A plot showing one category for each variable

119. What is multiple correspondence analysis?


A) A correspondence analysis between two variables with more than two
categories each
B) A technique which allows summarization of more than two non-metric
variable with a small number of metric co-ordinates
C) A sequence of simple correspondence analyses using different models

Chapter 15

120. What is the difference between an explanatory and a confirmatory factor


analysis?
A) Exploratory analysis imposes constraints on which variable loads on
which factor
B) Confirmatory analysis imposes constraints on which variable loads on
which factor
C) It is merely a philosophical distinction

121. What are latent variables?


A) Variables measured with a large degree of error
B) Unmeasurable variables which can be proxied by a set of other manifest
variables
C) Unmeasurable variables which can be estimated through regression
analysis

122. What is the difference between the structural and the measurement
equation in a SEM (Structural Equation Modeling)?
A) The measurement equation relates latent construct to manifest
variables, the structural equation describes the overall relationships
among endogenous and exogenous variables, including latent variables
B) The structural equation relates latent construct to manifest variables,
the measurement equation describes the overall relationships among
endogenous and exogenous variables, including latent variables
C) There is no difference

123. What is the difference between path analysis and structural equation
model?
A) Path analysis has no latent variables
B) Path analysis does not allow for correlation between explanatory
variable
C) Path analysis has no dependent variable

124. What is the distinction between over-identification, just-identification


and under-identification?
A) Over-identified models have more observations than parameters and
allow model testing, just-identification provide a unique estimate, and in
under-identification there are more parameters to be estimated than
available observation and the model cannot be estimated
B) Under-identified models have more observations than parameters and
allow model testing, just-identification provide a unique estimate, and in
over-identification there are more parameters to be estimated than
available observation and the model cannot be estimated
C) . Over-identified models have more latent variables than manifest ones,
under-identified the opposite and in just-identified models the number of
manifest variables equals the number of latent variables

125. What is the purpose of goodness-of-fit statistics?


A) An evaluation of the performance of the estimated model in reproducing
the original observations
B) An evaluation of the performance of the estimated model in reproducing
the original data covariance matrix
C) An evaluation of the amount of variability explained by the structural
model compared to the measurement model

126. What does a chi-square test with a probability of 0.10 mean in a SEM
framework?
A) That the theoretical model is not rejected by the data at the 5%
significance level
B) That the theoretical model is rejected by the data at the 5% significance
level
C) That difference between the original and estimated covariance matrices
is 10%

127. What is the meaning of box and circles in path diagrams?


A) Boxes refer to exogenous variables, circles to endogenous variables
B) Boxes refer to manifest variables, circles to latent variables
C) Boxes refer to variables, circles to errors
Chapter 16

128. Which of the following is a problem of ordinary regression with a binary


dependent variable?
A) Model predictions are not bounded between the two binary values
B) The distribution of residuals is not normal
C) Both of the above

129. Which of the following is the model for a nominal dependent variable
with metric explanatory variables?
A) The regression model
B) Multinomial logistic regression
C) Logit model

130. Which of the following is the link function for logistic regression?
A) The logistic transformation
B) The logit transformation
C) The exponential transformation

131. What is the difference between logistic regression and the probit model?
A) In logistic regression residuals follow the logistic distribution, in probit
models they follow the normal distribution
B) In logistic regression residuals follow the logit distribution, in probit
models they follow the logistic distribution
C) Logistic regression deals with metric explanatory variables, in probit
model the explanatory variables are categorical

132. What is the difference between revealed and stated preferences?


A) Revealed preferences are based on the observation of actual choices,
stated preferences are based on responses to a questionnaire
B) Revealed preferences are measured with metric scales, stated
preferences are categorical
C) There is no difference

133. What is conjoint analysis?


A) A statistical method for estimating non-linear relationships
B) A statistical approach to creating choice sets for a stated preference
analysis
C) A statistical method for relating metric variables to categorical variables

134. Which of the following is not a characteristic of conjoint analysis?


A) It is a compositional approach
B) It can be modelled through multinomial logit models
C) It is based on non-probabilistic samples

135. What is the difference between the GLM and the generalized linear
model?
A) The generalized linear model has a categorical dependent variable, the
GLM has a metric dependent variable
B) The generalized linear model has only categorical explanatory variables,
the GLM has only metric dependent variables
C) Both of the above

136. What is the difference between the ordered logit model and the
multinomial logit model?
A) The ordered probit model has a categorical ordered dependent variable,
the multinomial logit a metric continuous dependent variable
B) The ordered probit model has a categorical ordered dependent variable,
the multinomial logit a categorical nominal dependent variable
C) The ordered probit model has a categorical ordered dependent variable,
the multinomial logit has more than one categorical dependent variable

Vous aimerez peut-être aussi