Question 1

A researcher computes a null model then a random intercept fixed slope model. The researcher finds a BIC value of 5. Which one of the following is suggested?&#10;A) weak evidence&#10;B) positive evidence&#10;C) strong evidence&#10;D) very strong evidence

Accepted Answer

BIC value of 2-6 indicates positive evidence to favor model 2 over model 1.

Question 2

Which one of the following is the most important factor to consider when detecting cross-level interaction effects?&#10;A) average level 1 sample size&#10;B) average level 2 sample size&#10;C) variability of the level 1 slope coefficients&#10;D) within-group sample size

Accepted Answer

The power to detect cross-level interaction effects is driven by the within-group sample size (Spybrook, 2014).

Question 3

A researcher wishes to determine the reduction of within-group variability after adding predictors. Which one of the following should the researcher compute?&#10;A) Effect sizes for the between-group gamma coefficients for the intercepts&#10;B) Effect sizes for the individual beta coefficients within-groups&#10;C) proportion reduction in variance explained between-groups&#10;D) proportion reduction in variance explained within-groups

Accepted Answer

The proportion reduction in variance explained within-groups (i.e., at level 1) answers the question: How much is the within-group variability reduced after adding predictors? This effect is estimated as a pseudo-variance explained effect.

Question 4

A researcher finds an intraclass correlation coefficient of .10. Which one of the following reflects a correct interpretation of this?&#10;A) The ICC is statistically significant at an alpha of .50&#10;B) The probability of rejecting the null hypothesis when it is true is 10%&#10;C) The proportion of variation in the outcome that is between groups is 10%&#10;D) The within-group variation in the outcome is 10%

Accepted Answer

The ICC, by definition, is the proportion of variance in the outcome that is between higher level units; in this case, an ICC of .10 indicates 10% of the proportion of variance in the outcome is between higher level units

Question 5

A researcher examines the influence of self-efficacy on academic performance for students within classrooms. The results report an alpha of .01 and a p value of .05 for the slope variance for self-efficacy. Which one of the following is a correct interpretation of this?

A) The proportion of variation in academic performance that is between classrooms is 5%
B) The proportion of variance in the self-efficacy that is between classrooms is 1%
C) There is statistically significant variation across classrooms in the relationship between self-efficacy and academic performance
D) There is statistically significant variation within schools in academic performance as it relates to self-efficacy

Accepted Answer

The p is less than alpha and thus suggests that the slopes differ statistically significantly across classrooms.

Question 6

At which one of the following is sample size most important?&#10;A) all levels&#10;B) highest level&#10;C) level 1&#10;D) within-groups

Accepted Answer

In general, in MLM, the sample size at the highest level is primarily of most concern because the sample size at that level is always smaller than at the lowest level (Maas & Hox, 2005).

Question 7

A researcher finds statistically significant results for a deviance test. Which one of the following is suggested?&#10;A) Better model fit is evidenced for the more parameterized model&#10;B) Evidence to favor model 1 over model 2&#10;C) There is sufficient power&#10;D) There is statistically significant variation within groups

Accepted Answer

The deviance test follows a chi-square distribution, and statistically significant results indicate evidence that the more parameterized model has better model fit.

Question 8

Which one of the following centering methods is most appropriate when the researcher wishes to examine the extent to which there is a cross-level interaction between a group and a unit-level predictor?

A) Group mean centering
B) Grand mean centering
C) Either group or grand mean centering
D) Neither group nor grand mean centering

Accepted Answer

When interaction effects (aka moderating effects or cross-level interactions) are the primary concern, group mean centering is recommended as it offers a "pure" estimate of the moderation of the level 2 predictor on the level 1 relationship between the level 1 predictor and the outcome (Enders & Tofighi, 2007).

Question 9

Which one of the following should guide the choice in the centering method selected?&#10;A) the proportion of variation in the outcome that is between groups&#10;B) the research question&#10;C) the sample size at the highest level&#10;D) the sample size at the lowest level

Accepted Answer

There is not a necessarily correct versus incorrect choice in centering, but the selection needs to be based on the research question that is attempting to be addressed (Enders & Tofighi, 2007).

Question 10

A researcher examines the influence of self-efficacy on academic performance for students within classrooms. The results report an alpha of .01 and a p value of .05 for the slope variance for self-efficacy. Which one of the following is a correct interpretation of this?

A) The proportion of variation in academic performance that is between classrooms is 5%
B) The proportion of variance in the self-efficacy that is between classrooms is 1%
C) There is statistically significant variation across classrooms in the relationship between self-efficacy and academic performance
D) There is statistically significant variation within schools in academic performance as it relates to self-efficacy

Accepted Answer

With a random effect for the slope, the effect of the level 1 predictor (i.e., self-efficacy) varies randomly across groups (i.e., classrooms)

Deck 11: Multilevel Linear Modeling