Answers to Even-Numbered Exercises

Applied Multivariate Statistical Concepts, 1e Debbie Hahs- Vaughn (Solutions Manual All Chapters)

(Download link at the end of this file)

• / 4

Answers to Even-Numbered Exercises

Applied Multivariate Statistical Concepts, 1e Debbie Hahs-Vaughn

Chapter 1 Answers to Conceptual Problems

• d—structural equation modeling allows the examination of relationships or prediction

with multiple dependent variables

• a—logistic regression allows the examination of one categorical dependent variable

• c—multiple linear regression provides for examining a continuous outcome

• b—MANOVA is used when there are multiple outcomes with the goal of determining

mean differences

• c—multilevel linear modeling is used to analyze relationships of units nested within

groups (e.g., children within preschool)

• a—cluster analysis is a statistical technique for developing profiles of units (e.g., people)

• d—confirmatory factor analysis allows for grouping of constructs when there is strong

theoretical evidence to support the relationships between variables

• b—discriminant analysis is a technique to predict group membership

• d—propensity score matching allows units to be matched; which the matched groups can

then be used for later inferential analyses

• c—MANOVA is a test of mean differences, applicable when there are two or more

dependent variables

Chapter 2 Answers to Conceptual Problems

• c—see definition
• / 4

• a—cannot make Type II error there

• c—d is an effect size index, a measure of practical significance

• c—of these options, the best to visually examine the relationship between two variables is

a scatterplot

• c—equal sample sizes is not required

• b—pre to post mean differences is best examined using a dependent t test

• c—the intercept is 37000, which represents average salary when cumulative GPA is zero

• d—linear relationships are best represented by a straight line, although all of the points

need not fall on the line

• d—null hypothesis does not consider SS values

• false—with two groups the results of both procedures will always be the same

Chapter 3 Answers to Conceptual Problems

• false—the assumption of independence is the assumption that is most closely aligned

with the sampling design

• c—independence can be viewed with a scatterplot of residuals to predicted; random

display of points indicates evidence of the assumption being met

• d—kurtosis statistic that is within an absolute value of 7 is one form of evidence of

normality

• b—nonstatistically significant Box’s M is evidence that homogeneity of variance-

covariance has been met

• c—the same plots used to screen for independence can be used to screen for homogeneity

of variance; a scatterplot of residuals to predicted; random display of points indicates evidence of the assumption being met

• c—a relatively normal distribution will be evident when skewness is within an absolute

value of 2.0 and kurtosis within an absolute value of 7.0

• c—homoscedasticity applies to multiple linear regression, and assume the variation in

scores for one continuous variable is approximately equal to the variation in scores for 3 / 4

another continuous variable

• a—when all pairs of dependent variables are bivariate normally distributed, linearity

evidence is present

• d—VIF value of 30 suggests multicollinearity

• b—independence is generally met when there is simple random sampling

Chapter 4 Answers to Conceptual Problems

• b—partial correlations correlate two variables while holding constant a third

• a—as variable 3 has the largest correlation with variable 1 and the smallest with variable

2

• c—perfect prediction when the standard error = 0

• false—the intercept can be any value

• false—adding an additional predictor can result in the same R

2

• false—best prediction is when there is a high correlation of the predictors with the

dependent variable and low correlations among the predictors

• no—R

2 is higher when the predictors are uncorrelated

• no—the partial correlation may be larger than, the same as, or smaller than .6

• c—given there is theoretical support, the best method of selection is hierarchical

regression

• no—as discussed, these methods may yield different final models

• no—the purpose of the adjustment is to take the number of predictors into account; thus2

adj R may actually be smaller for the most predictors

• true—that is precisely the situation when we should be the most concerned about

collinearity

Answers to Computational Problems

• / 4