Deck 15: Using Multivariate Design and Analysis

Question 1

A design in which only one dependent variable is analyzed, or where multiple dependent variables are analyzed in separate statistical analyses, follows a _____ strategy.

A) single-variable
B) multivariate
C) univariate
D) unilevel

Answer 1

univariate

Question 2

When employing a multivariate strategy, you analyze:

A) two or more dependent variables in a single analysis.
B) several independent variables in a single analysis.
C) several independent variables but only one dependent variable in a single analysis.
D) None of the answers is correct.

Answer 2

two or more dependent variables in a single analysis.

Question 3

Which of the following is a multivariate statistical technique used to analyze correlational data?

A) Multiple regression
B) Discriminant analysis
C) Canonical correlation
D) All of the answers are correct.

Answer 3

All of the answers are correct.

Question 4

_____ is a variation of multiple regression in which the dependent or criterion variable is measured nominally.

A) Partial correlation
B) Discriminant analysis
C) Canonical correlation
D) Factor analysis

Question 5

An advantage of single multivariate analysis over multiple bivariate analyses is that:

A) you reduce the risk of finding relationships that occur merely by chance.
B) it is easier to compute than bivariate statistical techniques.
C) it allows you to look at only one relationship at a time.
D) None of the answers is correct.

Question 6

Which of the following is an advantage of a multivariate experimental strategy over a univariate strategy?

A) Collecting several dependent measures and treating them as a correlated set may reveal relationships that might be missed in a traditional univariate approach.
B) It tends to be more powerful than separate univariate designs.
C) Complex relationships among variables can be studied with less chance of making a Type I error.
D) All of the answers are correct.

Question 7

An assumption underlying bivariate correlational statistics is that the relationship between continuously measured variables is:

A) bidirectional.
B) linear.
C) nonlinear.
D) asymptotic.

Question 8

A _____ is a deviant score on a combination of variables from a single source.

A) univariate outlier
B) multivariate outlier
C) compound outlier
D) sampling anomaly

Question 9

Outliers can be handled:

A) by deleting outlying cases.
B) by transforming data.
C) either by deleting cases or variables or by using an appropriate data transformation.
D) None of the answers is correct.

Question 10

According to Tabachnick and Fidell (2013), data with a moderate positive skew should be transformed with _____.

A) a square root transformation
B) a logarithmic transformation
C) a reflecting strategy
D) a strategy that corrects data-entry errors

Question 11

When dealing with outliers, identify the first step in reflecting.

A) To subtract each score from the highest score in a distribution
B) To apply either a square root or logarithmic transformation
C) To check for transcription and other data-entry errors
D) To delete from the analysis all data from the variable containing the outlier

Question 12

According to Tabachnick and Fidell (2013), _____ is related to normality and is the assumption that "the variability in scores on one variable is roughly the same at all values of the other variable."

A) linearity
B) homoscedasticity
C) multicollinearity
D) canonicity

Question 13

_____ results when variables in your analysis are highly correlated.

A) Multicollinearity
B) Homoscedasticity
C) Heteroscedasticity
D) Normality

Question 14

The conical pattern of data points on a scatter plot indicates that _____ is present.

A) homoscedasticity
B) normality
C) heteroscedasticity
D) an outlier

Question 15

In the real world of research, the observed value of a variable _____ the true value of that variable.

A) is the same as
B) is always higher than
C) is always lower than
D) usually differs from

Question 16

The difference between the true value of a variable and the observed variable is the:

A) standard error.
B) error of measurement.
C) Type III error.
D) alpha error.

Question 17

_____ can arise from many sources, including inaccurate, incomplete, or biased sources of information.

A) Homoscedasticity
B) Type III error
C) Measurement error
D) All of the answers are correct.

Question 18

One good way to deal with the problem of measurement error is to:

A) use multiple sources of information for predictor variables.
B) statistically control error after your research has been conducted.
C) ignore it because it poses little problem for multivariate statistics.
D) None of the answers is correct.

Question 19

Which of the following statements is true of multivariate statistics?

A) Fewer subjects are needed for multivariate statistics than for univariate or bivariate statistics.
B) They are unaffected by the violation of assumptions that underlie them.
C) Outliers do not pose a major problem for multivariate statistics.
D) Fairly large sample sizes are needed for multivariate statistics.

Question 20

In _____, several dependent variables are analyzed to find out if any of them share common underlying dimensions.

A) factor analysis
B) canonical correlation
C) part correlation
D) discriminant analysis

Question 21

In factor analysis, each _____ is the correlation between a measure and an underlying factor.

A) beta weight
B) discriminant function
C) factor loading
D) structure correlation

Question 22

By convention, loadings are interpreted only if they are equal to or exceed:

A) .60.
B) .50.
C) .40.
D) .30.

Question 23

The strength of a factor is indicated by its:

A) eigenvalue.
B) factor loading.
C) structure correlation.
D) autocorrelation.

Question 24

Factor rotation is used to:

A) make the factors distinct.
B) reduce the amount of error variance in your data.
C) reduce the impact of outliers.
D) normalize data that are not normal.

Question 25

In _____ rotation, the axes representing the factors remain perpendicular when rotated around fixed points representing your data.

A) orthogonal
B) oblique
C) parallel
D) subtended

Question 26

The most popular orthogonal rotation method in factor analysis is:

A) oblique.
B) parallel.
C) varimax.
D) minimax.

Question 27

In the context of factor analysis, identify a true statement about oblique rotation.

A) It assumes that the measures and factors are correlated.
B) In this rotation, results are easier to interpret.
C) Its most popular method maximizes the variance of loadings on each factor and simplifies factors.
D) In this rotation, the axes representing the factors remain perpendicular when rotated around fixed points representing data.

Question 28

In principal components analysis, the diagonal of a completed correlation matrix is filled with _____.

A) communalities
B) rotated factor loadings
C) eigenvalues
D) ones

Question 29

Principal factors analysis completes a correlation matrix with _____ along the diagonal.

A) communalities
B) rotated factor loadings
C) eigenvalues
D) ones

Question 30

A(n) _____ is a measure of a variable's reliability.

A) eigenvalue
B) factor loading
C) communality
D) beta weight

Question 31

According to Tabachnick and Fidell (2013), identify a true statement about principal factors analysis.

A) It is most appropriate when one wants to reduce a large number of variables down to a smaller set.
B) It is most appropriate when one has theoretically or empirically based predictions.
C) It is often used in the early stages of research to identify the variables that cluster together.
D) It is used in the later stages of research where one can specify how variables relate given an underlying psychological process.

Question 32

_____ allows you to examine the relationship between two variables with the effect of a third variable removed from both of these variables.

A) Partial correlation
B) Part correlation
C) Factor analysis
D) Multiple regression

Question 33

_____ is used when you have a large set of variables that you want to describe in simpler terms and you have no a priori ideas about which variables will cluster together.

A) Confirmatory factor analysis
B) Exploratory factor analysis
C) Principle components analysis
D) Varimax rotation of factors

Question 34

_____ is used in situations where you may want to examine the relationship between two variables when the influence of a third is removed from only one of these variables.

A) Partial correlation
B) Part correlation
C) Factor analysis
D) Multiple regression

Question 35

In the context of the types of regression analysis, identify a true statement about simple regression analysis.

A) It is used when there is a well-developed theory or model suggesting a certain causal sequence of variables.
B) Sampling and measurement error tend to make unstable correlations among variables in this regression.
C) The order in which variables are entered is based on a statistical decision.
D) Each predictor variable is assessed as if it had been entered after each of the other predictors had been entered.

Question 36

In _____ regression, you specify the order in which variables are entered into the regression equation.

A) simple
B) stepwise
C) hierarchical
D) partial

Question 37

In the absence of any well-specified theory, you should usually choose:

A) hierarchical regression.
B) simple regression.
C) stepwise regression.
D) All of the answers are correct.

Question 38

Stepwise regression is used infrequently because it:

A) tends to capitalize on chance.
B) reduces error variance to an unacceptable level.
C) violates most of the assumptions underlying multivariate statistics.
D) None of the answers is correct.

Question 39

_____ is the correlation between the predicted values of Y and the observed values of Y.

A) R-square
B) Beta weight
C) Factor loading
D) Multiple R

Question 40

According to Tabachnick and Fidell (2013), an alternative to using beta weights to determine the unique contribution of each predictor is to use _____ instead.

A) factor loadings
B) structure correlations
C) squared semipartial correlations
D) R-square

Question 41

_____ is used when your dependent variable is categorical and you have several predictor variables.

A) Multiple regression
B) Canonical correlation
C) Factor analysis
D) Discriminant analysis

Question 42

The _____ is the average of the discriminant function scores calculated within a group.

A) multivariate mean
B) geometric mean
C) centroid
D) None of the answers is correct.

Question 43

In the context of discriminant analysis, the number of discriminant functions extracted is limited to the number of:

A) predictors.
B) levels of the dependent variable minus 1.
C) predictors or to the number of levels of the dependent variable minus 1, whichever is smaller.
D) dependent variables.

Question 44

Which of the following is a purely descriptive strategy and fails to infer causal relationships?

A) Discriminant analysis
B) Canonical correlation
C) Factor analysis
D) Part correlation

Question 45

The relationship between a set of predictor variables and a set of dependent variables can be determined using:

A) multiple regression.
B) discriminant analysis.
C) factor analysis.
D) canonical correlation.

Question 46

MANOVA operates by:

A) treating each of the several dependent variables separately.
B) forming new linear combinations of independent variables.
C) forming a new linear combination of dependent variables for each effect in your design.
D) None of the answers is correct.

Question 47

Research situations in which categorical variables are measured or manipulated are most effectively analyzed with:

A) MANOVA.
B) ANOVA.
C) factor analysis.
D) loglinear analysis.

Question 48

Loglinear analysis is analogous to:

A) chi-square.
B) MANOVA.
C) canonical correlation.
D) factor analysis.

Question 49

Loglinear analysis can be used:

A) for testing and building theoretical models.
B) if you conduct a correlational study with several categorical variables.
C) if your data violate the assumptions of parametric statistics (such as ANOVA).
D) All of the answers are correct.

Question 50

_____ is an application of multiple regression techniques to the testing of causal models.

A) Causal analysis
B) Path analysis
C) Factor analysis
D) Directionality analysis

Question 51

Path analysis:

A) can be used even in the absence of a well-defined causal model.
B) is a special case of MANOVA applied to correlational data.
C) should be used only if you have a clearly articulated causal model to test.
D) should be used to sift through data for interesting relationships among variables.

Question 52

Which of the following statements is true of exogenous variables?

A) They begin a causal sequence.
B) No causal paths lead to these variables.
C) They can sometimes be correlated.
D) All of the answers are correct.

Question 53

Which of the following statements is true about endogenous variables?

A) They begin the causal sequence in a path model.
B) Changes in them are believed to be caused by other variables.
C) They are external to a path model.
D) Measuring them directly is difficult because they are hypothetical constructs.

Question 54

In a(n) _____ model, there are no causal loops of variables.

A) exogenous
B) nonrecursive
C) recursive
D) open

Question 55

A(n) _____ consists of a sequence of events that eventually leads to variation in a dependent variable.

A) causal chain
B) open model
C) exogenous model
D) endogenous model

Question 56

_____ are used as estimates of the magnitude of causal relationships among variables.

A) Structure correlations
B) Path coefficients
C) Factor loadings
D) Discriminant functions

Question 57

Wright's rules are used to:

A) decompose paths into indirect effects.
B) determine the degree of causal connection between different exogenous variables.
C) determine the degree of causal connection between different endogenous variables.
D) determine the degree of causal connection between endogenous and exogenous variables.

Question 58

In structural equation modeling (SEM), variables in a model that are not directly observable are called _____ variables.

A) exogenous
B) endogenous
C) latent
D) manifest

Question 59

Which of the following is a confirmatory application of structural equation modeling (SEM)?

A) Strictly confirmatory approach
B) Alternative models approach
C) Model development approach
D) All of the answers are correct.

Question 60

When considering the correlations among dependent variables, multivariate statistical tests tend to be more powerful than separate univariate tests.

Question 61

Outliers affect both the magnitude and the sign of a calculated correlation.

Question 62

On a scatter plot, an elliptical pattern of data points indicates the presence of heteroscedasticity.

Question 63

To avoid the problem of error of measurement, you should develop a well-defined measurement model.

Question 64

Multivariate statistics are especially suited to research with small samples.

Question 65

In most cases, the results of a factor analysis are easier to interpret without rotation of factors.

Question 66

Oblique rotation assumes that measures are uncorrelated and consequently that the factors extracted are uncorrelated.

Question 67

Principal components analysis is most appropriate when you have theoretically or empirically based predictions.

Question 68

For most applications in psychological research, you should use the raw score regression weights to interpret a multiple regression analysis.

Question 69

A beta weight for a given predictor variable may be high because it directly produces most of the variance in a dependent variable.

Question 70

An alternative to using beta weights to determine the unique contribution of each predictor is to use squared semipartial correlations instead.

Question 71

R-square provides an index of the amount of variability in the dependent variable accounted for by the predictor variables.

Question 72

The Roy-Bargman test should be used in the absence of a theoretically or empirically based order of entry.

Question 73

Multivariate analysis of variance (MANOVA) circumvents problems with the homogeneity assumptions and error-term selection and is used for within-subjects designs.

Question 74

Curved arrows in a path model indicate causal relationships.

Question 75

Variables that begin a causal chain are called endogenous variables.

Question 76

Path analysis is designed to sift through data for interesting relationships among variables.

Question 77

Structural equation modeling (SEM) is a variant of path analysis.

Question 78

The values of measured variables fail to determine the value of a latent variable.

Question 79

Structural equation modeling (SEM) is normally used as an exploratory procedure.

Question 80

In structural equation modeling (SEM), researchers can evaluate the relationships among measured variables and hypothetical constructs within the models that they test.