Gender Bias in Prediction of Graduate Grade Performance from Miller Analogies Test Scores

J. DANIEL HOUSE Office of Institutional Research Northern Illinois University

EDWARD J. KEELEY Office of Institutional Research Loyola University

ALTHOUGH THE MILLER ANALOGIES TEST (MAT) is widely used as an admissions test for graduate programs, no published studies have investigated whether MAT scores predict subsequent achievement similarly for all subgroups of graduate students. When systematic differences exist in the relationship between test scores and performance on some criterion measure for various subgroups of students, then prediction bias exists ( Reynolds, 1982). A number of researchers have investigated gender bias in the prediction of student grade performance from admissions test scores. In general, the results of these studies have indicated that admissions test scores tend to underpredict the subsequent

-353-