skip to main content skip to footer

An Intragroup Study of Similarities and Differences in the Prediction of GRE General Test Scores

Author(s):
Holland, Paul W.; Hoffman, Karla; Thompson, Martha J.
Publication Year:
2002
Report Number:
RM-02-02
Source:
ETS Research Memorandum
Document Type:
Report
Page Count:
187
Subject/Key Words:
Education Level, Gender, Multiple Regression Analysis, Parents' Education, Undergraduate Major, Racial and Ethnic Groups

Abstract

The authors analyzed the data for approximately 1 million Graduate Record Examination (GRE) tests taken from 1995 to 1999 to study the relationships between several items in the GRE Background Information Questions (BIQ) and GRE scores within 18 combinations of gender and race/ethnicity categories. The study is intragroup in the sense that we use the BIQ variables to predict GRE scores within each of these groups rather than comparing the GRE scores across these groups. The authors find that undergraduate major, reasons for taking the GRE, undergraduate GPA, and education level at the time of test taking to be consistently important predictors across all of the groups we studied. Other factors that were important but less consistently important across all groups were graduate major, full-time or part-time graduate study, parent's education, and English language proficiency. Age-at-testing and years-since-the-undergraduate-degree had no significant relationship with GRE scores for any of the groups. The groups varied somewhat in the degree to which GRE scores were predictable from these background variables, with the multiple regression R-squares ranging from about 17% to 37%. There was considerable consistency across the groups, with several interesting exceptions, in the BIQ categories of the test takers who would be predicted from the regression models to have the highest scores on the GRE.

Read More