Gender Differences in Prediction of Graduate Course Performance from Admissions Test Scores: An Empirical Example of Statistical Methods for Investigating Prediction Bias. AIR 1998 Annual Forum Paper.

This study focused on gender differences in examining the extent to which Graduate Record Examination (GRE) scores predicted subsequent achievement. Data on 275 graduate students in professional psychology programs at a large midwestern university were collected and analyzed. Two methods for the identification of prediction bias were used and student performance in 10 specific graduate courses was examined. It was found that GRE scores were effective predictors of the graduate course performance of the students. However, for a number of courses, GRE scores under predicted the achievement of female students and over predicted the achievement of male students. Specifically, GRE verbal scores were found to under predict female achievement in four courses, while GRE quantitative scores under predicted female achievement in seven courses. GRE analytical scores under predicted female achievement in four courses, while overall GRE scores under predicted female achievement in six courses. (Contains 48 references.) (MDM)