Statistics Module 8: Chapter 15

The difference between ANCOVA and MANCOVA is that a) with a MANCOVA, variables can be partialed out; with an ANCOVA, they cannot b) with an ANCOVA, variables can be partialed out; with a MANCOVA, they cannot c) a MANCOVA can have more than one dependent variable; an ANCOVA cannot d) a MANCOVA can have more than one independent variable; an ANCOVA cannot

A MANCOVA can have more than one dependent variable; an ANCOVA cannot

Test reliabilities are most likely to be discussed in research articles when a) a new measure is developed b) results of the study are significant c) reporting results of a chi-square test d) background variables, such as age and sex, are being considered

A new measure is developed

An investigator wants to know the relation between burden and mastery over and above any possible effects of depression. The procedure the investigator would use is a) bivariate correlation b) partial correlation c) factor loading d) factor analysis

Partial correlation

The degree of association between two variables, over and above the influence of one or more other variables, is a) path analysis b) reliability c) partial correlation d) factor analysis

Partial correlation

If the results of a study are presented as a diagram with arrows connecting variables, the analysis is a) factor analysis b) partial correlation c) path analysis d) multiple regression

Path analysis

If a measure has low reliability, the correlation between it and any other variable will typically be a) enlarged b) reduced c) enlarged if the correlation is positive and reduced if the correlation is negatived d) reduced if the correlation is positive and enlarged if the correlation is negative

Reduced

A social psychologist who correlates the responses of half of the items on a questionnaire with the other half to check the accuracy of the measure is assessing a) a model's fit b) test-retest reliability c) split-half reliability d) interrater reliability

Split-half reliability

A covariate is best described as a) a dependent variable that is correlated with the independent variables b) a variable that is controlled or partialed out c) a variable that is thought to be an effect of all other variables measured. d) a variable that does not seem to have any effect on the others

A variable that is controlled or partialed out

The difference among means of groups is ordinarily tested by a) multiple correlations and multiple regressions b) multiple regressions and bivariate predictions c) analysis of variance and t tests d) t tests and bivariate correlations

Analysis of variance and t tests

An advantage of structural equation modeling over ordinary path analysis is that structural equation modeling a) does not require significance tests b) analyzes the links among latent variables c) analyzes the links among measured variables d) can handle equal-interval variables

Analyzes the links among latent variables

When ratings are made on a nominal scale, interrater reliability should be assessed using a) a Pearson correlation b) Spearman's rho c) Cohen's kappa d) Cronbach's alpha

Cohen's kappa

The final step of a mediational analysis tests whether the association of X and Y a) increases when the mediator is taken into account b) decreases when the mediator is taken into account c) is larger than the association of X with the mediational variable d) is smaller than the association of X with the mediational variable

Decreases when the mediator is taken into account

A limitation of causal modeling is that a) even when all predicted paths are significant, there is an overall good fit, and the significance test is as expected, it is possible that there are other patterns of causality that work better b) only the cause is examined, and comparisons and linkages between the different variables are impossible c) it cannot be done very easily on a computer d) path diagrams are rarely accurate and often give a false picture, tending to suggest that the true linkages are among the measured variables, whereas they are usually among the latent variables

Even when all predicted paths are significant, there is an overall good fit, and the significance test is as expected, it is possible that there are other patterns of causality that work better

A developmental psychologist has administered a questionnaire to a large group of fathers to see how frequently they participate in each of 40 activities with their children. The psychologist wants to know if there are some activities that can be grouped together such that if a father does one activity, he is likely to do the other, but in which there is little association with activities in other groupings. The appropriate procedure is a) partial correlation b) analysis of covariance c) factor analysis d) path analysis

Factor analysis

A social psychologist has found 20 different variables that seem to affect willingness to take risks and wants to simplify these results by grouping them. The appropriate procedure is a) partial correlation b) test-retest reliability c) factor analysis d) multivariate analysis of variance

Factor analysis

The association of a variable to a factor is called the variable's a) factor analysis b) factor loading c) partial factor d) path factor

Factor loading

A general formula whose underlying logic connects t tests, analysis of variance, correlation, and multiple regression is the a) nonparametric model b) general linear model c) factorial model d) general factor analytic model

General linear model

A test would be considered most useful when its reliability is a) less than .10 b) less than .40 c) greater than .60 d) greater than .90

Greater than .90

When reading a research article based on an unfamiliar statistical procedure, a basic understanding may be obtained by a) seeking some indication of the q coefficient or, if that is not available, some indication of power b) identifying the significance level and the effect size c) following the path diagram or contingency table, depending on the level of measurement of the variables involved d) plotting the regression line

Identifying the significance level and the effect size

In the general linear model, errors in predictions are caused by a) influences of measured predictor variables b) influences not measured by any of the predictor variables c) some fixed influence that is the same for all individuals d) correlations among the predictor variables

Influences not measured by any of the predictor variables

The purpose of a mediational analysis is to examine whether a presumed causal relationship between two variables a) is due to some intervening variable b) is different according to the level of some additional variable c) is caused by some third variable d) might be causal in the opposite direction from the initial assumption

Is due to some intervening variable

If an investigator using structural equation modeling measures school status as a combination of GPA, number of activities, and involvement in school government, school status is a a) latent variable b) manifest variable c) measurement variable d) causal variable

Latent variable

The variables about the people or observations within each grouping in a multilevel model are a) upper-level variables b) middle-level variables c) lower-level variables d) criterion-level variables

Lower-level variables

An advanced type of regression analysis that handles research situations in which people are grouped in some way that could affect the pattern of scores is a) multilevel modeling b) multiple regression c) partial correlation d) factor analysis

Multilevel modeling

Hierarchical linear modeling is a common name for a) path analysis b) factor analysis c) multilevel modeling d) structural equation modeling

Multilevel modeling

The formula Ŷ = a + (b1)(X1) + (b2)(X2) + (b3)(X3) is the formula for a) multiple regression b) the analysis of variance c) bivariate correlation d) the t test

Multiple regression

If an investigator wants to be able to include more than one independent variable, more than one dependent variable, and wants to partial out a variable, the appropriate procedure is a) factor analysis b) analysis of covariance (ANCOVA) c) multivariate analysis of variance (MANOVA) d) multivariate analysis of covariance (MANCOVA)

Multivariate analysis of covariance (MANCOVA)

In structural equation modeling, a) for most fit indexes, the closer to -1, the better the fit. b) on the RMSEA, the closer to 0, the better the fit c) a significant result means a better fit than a nonsignificant result d) for most fit indexes, the closer to 0, the better the fit

On the RMSEA, the closer to 0, the better the fit

A technique that uses a path diagram with arrows between variables, but that has some advantages over path analysis is a) partial correlation b) multiple regression c) factor analysis d) structural equation modeling

Structural equation modeling

An educational psychologist who gives the same test twice to the same people to check the accuracy of the test is assessing a) a model's fit b) test-retest reliability c) split-half reliability d) interrater reliability

Test-retest reliability

The t test can be derived from all of the following EXCEPT a) multiple regression b) the Z test c) bivariate correlation d) the analysis of variance

The Z test

Cronbach's alpha can be thought of as a) the average of the correlations between two halves of a test based on every possible split of the items into two halves b) the average of the largest and smallest correlations among test items c) the square root of the standardized path coefficient d) the standardized path coefficient squared

The average of the correlations between two halves of a test based on every possible split of the items into two halves

In the formula Y = a + (b1)(X1) + (b2)(X2) + e, the term "b1" refers to a) the fixed influence that applies to all individuals b) the sum of influence of all variables on the first criterion variable c) the degree of influence of the first predictor variable on the criterion variable d) the degree of error associated with the first predictor variable

The degree of influence of the first predictor variable on the criterion variable

The "least-squares criterion" sets the a's and b's in the general linear model so that a) the sum of e2 over all participants is 0 b) the sum of e2 over all participants is 1 c) the sum of e2 over all participants is as low as possible d) the sum of e2 over all participants is as high as possible

The sum of e2 over all participants is as low as possible

The general linear model is called a "linear model" because a) a single-tracked line of reasoning drives the analysis b) the model makes predictions based on the intersection of two or more linear trends c) the computations rely on a curvilinear formula d) there are no squared (or higher power) terms in the equation

There are no squared (or higher power) terms in the equation

The goal of path analysis is a) to group variables b) to eliminate unnecessary variables c) to arrange variables from most to least influential d) to examine the causal structure of a group of variables

To examine the causal structure of a group of variables

A significant MANCOVA is usually followed up with a) ordinary t tests b) partial correlations c) one-way ANOVAs d) univariate ANCOVAs.

Univariate ANCOVAs.

The variables about the grouping as a whole in a multilevel model are a) upper-level variables b) middle-level variables c) lower-level variables d) criterion-level variables

Upper-level variables

The difference between MANCOVA and MANOVA is that a) a MANCOVA can have more than one dependent variable; a MANOVA cannot b) a MANCOVA can be done with a computer; a MANOVA cannot c) with a MANCOVA, variables can be partialed out d) with a MANOVA, variables can be partialed out

With a MANCOVA, variables can be partialed out

If an educational psychologist studying the effect of three different teaching programs on students' learning wants to eliminate the effect of initial differences in learning ability among the students, the psychologist should use a) multivariate analysis of variance (MANOVA) b) analysis of variance (ANOVA) c) analysis of covariance (ANCOVA) d) test-retest reliability

analysis of covariance (ANCOVA)