APPENDIX D

REGRESSION SUMMARY DETAILS

This Appendix provides detailed summary information from the regression analyses conducted on the Cooperative Institutional Research Program (CIRP) and Michigan Student Study (MSS) data bases. (The IGRCC results were analyzed using Analysis of Variance, and do not lend themselves to this presentation format.)

Regressions based on CIRP data

Four separate regression models were developed for each outcome considered in the CIRP data base, with the models differing only by the specific informal interaction measure considered in combination with the classroom diversity measure. For each of the four models, four columns of numerical information are presented. The first column, labeled r, indicates the raw (or zero-order) correlation between each of the predictor variables and the outcome measure, and is typically the upper-bound estimate of the independent effect that each diversity measure has on the outcome measure. Columns 2 (labeled Step 1), 3 (labeled Step 2), and 4 (labeled Step 3) indicate whether or not the predictor variable is still a significant predictor of the outcome variable after controlling for other variables in the model that might also explain the relationship between diversity experiences and student outcomes. For each of the analyses, Step 1 controls (or removes) the possible effects of student background characteristics and (where available) a student's response to the outcome measure upon entry into college; Step 2 adds the diversity experience measures to the prediction equation; Step 3 adds institutional characteristics such as structural diversity, institutional selectivity, and type (such as public versus private, four-year college versus university). Appendix C provides a list of measures for each of the steps of the analysis.

When a diversity experience measure remains statistically significant after controlling for all the other variables in the analysis, this provides strong evidence of the importance of this variable in explaining the relevant outcomes, as all of the competing explanations were given the better chance to explain the finding.

Given differences in sample sizes across racial/ethnic group, effects were judged as significant for White students at the p < .05 level (Table D1), while the criterion of p < .10 was used for the much smaller African American and Latino samples (Tables D2 and D3, respectively). Significant positive effects are indicated by a square symbol (), while significant negative effects are indicated by a diamond symbol (). Nonsignificant findings are not shown.

Regressions based on MSS data

Eight separate regression models were developed for each outcome considered in the MSS data base, with the models differing only by the specific informal interaction measure considered in combination with the classroom diversity measure. The format of the summary tables that follow is similar to that used for the CIRP analyses, but differs slightly due to differences in the analytical approach used.

For each of the eight models, three columns of numerical information are presented. The first column, labeled r, indicates the raw (or zero-order) correlation between each of the predictor variables and the outcome measure, and is typically the upper-bound estimate of the independent effect that each diversity measure has on the outcome measure. Columns 2 (labeled Step 1) and 3 (labeled Step 2) indicate whether or not the predictor variable is still a significant predictor of the outcome variable after controlling for other variables in the model that might also explain the relationship between diversity experiences and student outcomes. For each of the analyses, Step 1 controls (or removes) the possible effects of student background characteristics and (where available) a student's response to the outcome measure upon entry into college; Step 2 adds the diversity experience measures to the prediction equation. (The Step 3 column provided for the CIRP data is unavailable for the MSS analysis since it is based on data from a single institution, and as a result institutional characteristics do not vary). As before, when a diversity experience measure remains statistically significant after controlling for all the other variables in the analysis, this provides strong evidence of the importance of this variable in explaining the relevant outcomes, as all of the competing explanations were given the better chance to explain the finding.

Given differences in sample sizes across racial/ethnic group, effects were judged as significant for White students at the p < .05 level (Table D4), while the criterion of p < .10 was used for the much smaller African American sample (Tables D5). Significant positive effects are indicated by a square symbol (), while significant negative effects are indicated by a diamond symbol (). Nonsignificant findings are not shown.

Next "Gurin" Section	|	"Gurin" Table of Contents	|	"Compelling Need . . ." Table of Contents

Questions? Comments? Please send e-mail to diversitymatters@umich.edu. Site last updated: September 5, 2012.   Copyright © 1997–2013 Regents of the University of Michigan.