Chapter 17
Data Analysis: Investigation of Association
a. Correlation analysis involves the measurement of the closeness of the relationship
between two or more variables.
b. Regression analysis involves the derivation of an equation that relates the
criterion variable to one or more predictor variables.
c. Regression analysis can establish the causal relationship between two or more
variables.
d. The least squares equations result from a probabilistic model which relates the
variables.
e. The sum of the errors about the regression line in a least squares solution equals
zero.
n
a. r = Σ xiyi where xi and yi are deviations from the mean.
i=1
b. r = 1- sy/x.
sy
c. Unexplained variation
Total variation
Σ xiyi
d. r= _________
nsxsy
e. None of the above.
regression model must be satisfied?
a. The mean of the disturbance term must equal one.
b. The variance of the disturbance term is independent of the values of the predictor
variable.
c. The values of the error term are independent of one another.
d. all of the above.
e. Both b and c above must be satisfied.
squares regression?
a. The mean or average value of the error term is zero.
b. The variance of the error term is constant across levels of the predictor variable.
c. The values of the error term are independent of the predictor variable.
d. The values of the error term are independent of one another.
e. All of the above are necessary assumptions underlying the least squares solution.
01 we may
a. conclude with absolute certainty that no relation exists between x1 and y.
b. have simply made a Type II error, and thus a relation between x and y may exist.
c. have simply chosen the wrong form of the model to investigate, as the relation
between x1 and y may be nonlinear.
d. conclude that b and c above.
e. conclude that none of the above.
(Use the following information to answer the next two questions.)
Suppose that the relationship between sales (Y, in $000) and number of salespeople (X) is
represented by the following regression equation:
Y = 105.2 + 35.8X
a. $35.80
b. $141,000
c. $35,800
d. $141.00
e. more information is needed to answer this question
a. $358,000
b. $463.20
c. $358.00
d. $463,200
e. more information is needed to answer this question
Y2 = 20 – 39X
s2Y/X = 360 s2Y = 3600 r = .90
Which of the following statements is FALSE?
a. For every unit change in X there is a corresponding negative change in the
average value of Y of 39 units.
b. 90 percent of the variation in Y is associated with variation in X.
c. The estimated average variance of the deviations of the errors about the regression
line equals 360.
d. The average value of Y given x = 10 is -370.
e. They are all true.
made in the regression model is
a. the observations come from a bivariate normal distribution.
b. the variance of the error term is dependent on the values of the predictor variable.
c. the variance of the disturbance term is constant.
d. the values of the error term are dependent on each other.
e. the Xi are fixed.
is TRUE?
a. x and y are highly related, whereby a positive change in x is accompanied by a
positive change in y
b. the two variables x and y are not related to one another
c. x and y are highly related, whereby a negative change in x is accompanied by a
positive change in y
d. the coefficient of determination is equal to -.81
e. c and d above
^
statistic to determine if this value of ß could have been due to chance.
a. F = 17.57
b. t = 0.108
c. F = 9.30
d. t = 17.57
e. t = 9.30
determination?
Explained variation
a. r2 = 1- Total variation
Total variation – Unexplained variation
b. r2 = Total variation
Unexplained variation
c. r2 = 1- Total variation
Explained variation
d. r2 = Total variation
e. All of the above are formulas for computing the coefficient of determination.
^
a. is Y, which identifies the predictor variable.
b. is 1, which identifies the predictor variable, of which this ß-value is the
coefficient.
c. is 2, which indicates another predictor variable in the regression equation.
d. is 3, which indicates another predictor variable in the regression equation.
e. is Y, which identifies the criterion variable.
partial regression equation to be interpreted as the average change in the criterion
variable associated with a unit change in the appropriate predictor variable holding
other predictor variables constant?
a. The predictor variables must be correlated.
b. The variance among predictor variables must be equal.
c. The criterion variable must be normally distributed.
d. The predictor variables must be uncorrelated.
e. None of the above are necessary assumptions.
partial regression, ßY1.2 and ßY2.1, can be interpreted as
a. the unit change in the criterion variable associated with an average change in the
appropriate predictor variable while holding the other predictor variable constant.
b. the change in the criterion variable associated with an average change in the
predictor variables.
c. the average change in the criterion variable associated with an average change in
the appropriate predictor variable while holding the other predictor variable
constant.
d. the average change in the criterion variable associated with a unit change in the
appropriate predictor variable while holding the other predictor variable constant.
e. the average change in the criterion variable associated with a unit change in the
appropriate predictor variable.
a. the predictor variables are independent.
b. the criterion variables are independent.
c. the criterion variables are correlated among themselves.
d. the predictor variables are correlated among themselves.
e. the error terms are correlated.
^
^
a. Multicollinearity is said to be present in a multiple regression problem when the
predictor variables are correlated among themselves.
b. If the data are highly multicollinear, there will be a tendency to judge many of the
predictor variables as not being related to the criterion variable when in fact they
are.
c. The standard error of estimate of the least squares coefficients increases as the
dependence among the predictor variables increases.
d. A multicollinear condition within a data set increases the efficiency of the
estimates for the regression parameters.
e. Multicollinearity makes interpretation of the beta coefficients problematic.
a. suggests that 86 percent of the variation in Y is associated with variation in X2.
b. suggests that 86 percent of the variation in Y is explained by or is associated with
variation in X1 and X2.
c. suggests that 86 percent of the variation in Y not associated with X2 is
incrementally associated with X1.
d. suggests that 86 percent of the variation in Y not associated with X1 is
incrementally associated with X2.
e. suggests that 86 percent of the variation in Y is associated with variation in X1.
a least squares analysis. The estimate was based on 50 observations. How many
degrees of freedom are associated with this test?
a. 50
b. 48
c. 49
d. 47
e. none of the above
a particular product Y to X1 and X2, where
X1 X2
–If a person belongs to lower class 0 0
-If a person belongs to middle class 1 0
-If a person belongs to upper class 0 1
Which of the following is FALSE? The equation suggests:
a. An upper class person could be expected on the average to spend $38 per year
more than a lower class person on the product.
b. A middle class person could be expected on the average to spend $16 more per
year on the product than a lower class person.
c. A lower class person could be expected on the average to spend $5 per year on the
product.
d. An upper class person could be expected on the average to spend $22 more per
year on the product than a middle class person.
e. An upper class person could be expected on the average to spend $38 per year on
the product.
regression equation.
a. m + 1
b. (m + 1)/2
c. m – 1
d. (m – 1)1/2
e. (m – 1)/2
approval ratings given to a particular politician. The hypothesis was tested using
regression analysis. If the variable education had seven categories, how many dummy
variables are necessary to include this variable in a regression equation?
a. 1
b. 7
c. 6
d. 8
e. none of the above
a. can be estimated employing transformations and the simple linear regression
model.
b. cannot be estimated employing transformations and the simple linear regression
model.
c. requires the multiple regression model without transformations be employed to
estimate them.
d. requires the multiple regression model with transformations be employed to
estimate them.
e. cannot be estimated with regression analysis.
criterion variable y and a predictor variable having four categories?
a. y = α + ß1X1
b. y = α + ß1X1 + ß2X2
c. y = α + ß1X1 + ß2X2 + ß3X3
d. y = α + ß1X1 + ß2X2 + ß3X3 + ß4X4
e. none of the above accurately captures the situation
useful for which of the following reasons?
a. Transformations allow regression analysis when the relationship between
predictor and criterion is not additive.
b. Transformations allow regression analysis when the relationship between
predictor and criterion is nonlinear.
c. Transformations allow regression analysis when the error terms are
heteroscedastic.
d. None of the above are valid reasons for transforming predictor and criterion
variables.
e. All of the above are valid reasons for transforming predictor and criterion
variables.
The following questions focus on Conjoint Measurement:
a. requires subjects to make judgments about stimuli that represent some
predetermined combinations of attributes.
b. is like regression analysis with dummy variables.
c. relies on the ability of respondents to express their preferences for various
combinations of attributes.
d. a and b.
e. a, b, and c.
a. determine the value systems people are using in making choices from
combinations of attributes.
b. determine the perceived similarity of various objects.
c. determine the perceived similarity of various attributes.
d. measure the person’s preference for specific objects.
e. none of the above.
a. Subjects are asked to make judgments about stimuli that represent some
predetermined combinations of attributes.
b. Respondents are asked to indicate their preference for each of the stimuli.
c. While the attributes that will be used to construct the stimuli will stem
primarily from the purpose of the investigation, the analyst will have some
discretion in this regard.
d. The attributes used to construct the stimuli should be actionable for the
company.
e. Subjects are asked to make judgments about existing objects.
a. An advantage of conjoint analysis over regression is that the dependent
variable can also be a dummy variable.
b. In a conjoint analysis application, the basic aim is to determine the features of
products that respondents most prefer.
c. Of primary concern to researchers employing conjoint analysis are the trade-
offs consumers make when evaluating various combinations of product
attributes.
d. Conjoint analysis depends on preference judgments that are ratings or
rankings.
e. Conjoints are easily run via multiple regression.
a. requires respondents to express how they are weighting particular attributes in
forming their judgments for products.
b. uses actual products or brands as the typical stimuli.
c. gets its name from the fact that we can measure relative values of things
considered jointly which might be unmeasurable taken one at a time.
d. a and b.
e. a, b, and c.
a. combinations of attitudes.
b. products for which perceptions of similarity will be recorded.
c. predetermined combinations of product attributes.
d. groupings of related products.
e. none of the above.
a. It is necessary to present all possible combinations of attribute levels to
respondents in a conjoint analysis.
b. The attempt in a conjoint analysis solution is to assign values to the levels of
each of the attributes so that the resulting values or utilities are as monotonic as
possible with the input rank order judgments.
c. In a typical conjoint analysis, the utilities assigned to each attribute are
combined by multiplying them to determine the total utility for each
alternative.
d. a and b.
e. a, b, and c.
a. Conjoint analysis does not require the respondent to provide self-reports
regarding the importance of various product attributes.
b. Conjoint analysis attempts to determine individuals’ utilities for various
product attributes when determining choices.
c. Conjoint analysis provides essentially the same results as a multidimensional
scaling analysis in a given situation.
d. Conjoint analysis is quite dependent on the availability of a computer.
e. All of the above are true of conjoint analysis.
a. The attributes used to construct the stimuli for a conjoint analysis are
completely determined by the purpose of the investigation.
b. The attributes used to construct the stimuli for a conjoint analysis should be
both actionable for the company and important to people.
c. In a typical conjoint analysis study, most of the attributes that could be used to
construct stimuli will be used.
d. Our ability to generate good estimates of the utility of each attribute level
depends upon the number of stimuli being relatively small versus the number
of parameters that need to be estimated.
e. They are all false.
a. the respondent’s preference for each object or brand.
b. the respondent’s perception of each object or brand.
c. the respondent’s utility for each attribute or product feature.
d. the mapping of perceptions and preferences.
e. none of the above.
a. The typical stimuli are actual products or brands.
b. Respondents are asked to make judgments about the relative similarity of
objects.
c. The attributes used for the stimuli should be important to people.
d. Most of the attributes that could be used will be used.
e. All the above statements are false.
a. The utilities or part-worth functions for a single subject can be determined.
b. The utilities can be used to determine the relative importance of each of the
attributes.
c. The importance of any attribute is determined by the spread in utilities between
the highest and lowest rated levels of the attribute.
d. A subject’s preferences for any attribute must be monotonic for the procedure
to work.
e. The importance of an attribute to an individual can depend on the levels of the
attribute used when securing judgments.
development process.
a. idea generation
b. business analysis
c. concept evaluation
d. test marketing
e. none of the above
used in conjoint analysis?
a. managerial judgment
b. focus groups
c. analysis of insight-stimulating examples
d. depth interviews
e. All of the above methods can be used.
a. A linear preference model for a conjoint analysis study implies that subjects
always prefer more or less of the attribute.
b. The utilities of the various levels of an attribute used for conjoint analysis can
be used to infer the importance of an attribute to a subject.
c. An important trade-off occurs for respondents between easier judgments but
more of them when using the full profile approach to presenting attribute
combinations in a conjoint analysis.
d. It is generally recommended that the ranges for the various attributes in a
conjoint analysis study be made somewhat larger than what is normally found.
e. a and c.
a. 2.
b. 3.
c. 4.
d. 5.
e. more than 5.
order judgments in conjoint analysis is FALSE?
a. The rating scale method is currently more popular.
b. A fundamental premise underlying the rank order method is that it makes the
judgment task as close as possible to a consumer’s behavior while actually
shopping.
c. Rating scale data are easier to analyze.
d. When the rating method is used, subjects are asked to make relative judgments
with respect to their preference for one alternative versus another.
e. a and d.
a. The use of a linear preference model implies that subjects always prefer more
or less of the attribute.
b. If the analyst suspects that subjects might prefer an intermediate level of an
attribute, the analyst should attempt to fit their preferences using a part-worth
model.
c. The ranges of the various attributes should be somewhat narrower than what is
normally found so as to increase the believability of the task.
d. The number of stimuli should be relatively large versus the number of
parameters to be estimated.
e. With the paired-comparison data collection method, respondents can be asked
to simply choose which alternative in each pair they prefer.
presenting stimuli in conjoint analysis?
a. paragraph description
b. verbal description
c. prototype evaluation
d. pictorial representation
e. all of the above are typically used
a. The technique is restricted to product evaluations.
b. The utilities assigned each attribute level are typically added to determine the
total utility for each alternative.
c. The pairwise procedure treats two attributes at a time but considers all possible
pairs.
d. It is typically easier for subjects to supply pairwise judgments than full profile
judgments.
e. Subjects typically need to make more judgments with the pairwise approach
than the full profile approach.
a. the design of a new type of motorcycle
b. the determination of an appropriate price for a new electric fan
c. the determination of a market segment, or “niche,” for very high quality
electronic equipment
d. the selection of advertising media for a new product introduction
e. conjoint analysis is useful in each of these situations
a. The pairwise procedure in conjoint analysis considers all attributes at the same
time but only allows each attribute to be set at two or a pair of levels.
b. An individual typically needs to make more judgments when the full profile
method of data collection is used in conjoint analysis than when the pairwise
method is used.
c. The use of rating scales in lieu of rank order judgments to measure preference
in conjoint analysis has been increasing in popularity.
d. When the rating method is used in conjoint analysis, subjects are asked to
make relative judgments with respect to their preference for one alternative
versus another.
e. They are all false.
paired-comparison approach to determining attribute combinations in a conjoint
analysis?
a. The paired comparison approach works well with computer-interactive
interviewing procedures.
b. The paired comparison approach allows the researcher to check respondents’
consistency in their judgment of alternatives.
c. The paired comparison approach, although a more difficult task for
respondents than the full profile approach, is considerably more thorough.
d. All of the above are reasons.
e. b and c only are NOT reasons
Use the following information to answer the next 5 questions:
A marketing research team collected preferences from consumers in the MidWest who rated the
likelihood that they would go on each of the 16 possible week-long spa packages created by
combining the factors: San Diego, CA or Phoenix, AZ; $1499 or $2599; golf green fees included
or not; and daily massages included or not. (The factors were scored: 0=San Diego, 1=Phoenix;
0=$1499, 1=$2599; 0=golf not included, 1= golf included; 0 = massage not included, 1=
massage included. The likelihood scale ranged from 0 to 100, with 100 meaning greater
preference.) When the marketing research team conducted the conjoint analysis, they obtained
the following beta-weights (after averaging across the sample):
likelihood = -.85 destination -. 15 price +.70 golf +.50 massage
a. destination
b. price
c. golf
d. massage
e. golf with price, or destination
a. San Diego
b. Phoenix
c. Tucson (it’s in between)
d. Cancun would do better because it is closer to the MidWest
e. indeterminate, more analytical details are required to make that
specification
a. San Diego, by a lot
b. Phoenix, by a lot
c. San Diego, marginally
d. Phoenix, marginally
e. San Diego, only if they can golf
a. 15% discount
b. $1499
c. $2599
d. it depends on the segmentation structure
e. cheaper is always better
a. destination wouldn’t matter
b. consumers are price insensitive
c. the respondents weren’t golfers
d. the respondents don’t care for spas
e. the golf course could be exchanged for a spa