Testing the Predictive Validity of the GRE Exam on Communication Graduate Student Success: A Case Study at University at Buffalo

Abstract

GRE scores for entering M.A. and Ph.D. students in Communication at University at Buffalo from 1990 to 2001 were used to predict graduate student success. Graduate student success was measured in two ways: grade point average (GGPA) and graduation rates. Preliminary analyses on Master's students (M.A.) revealed that international students, compared to domestic students, earned lower GRE verbal (GRE-V) scores, higher GRE quantitative (GRE-Q) scores, had higher undergraduate grade point averages and were more likely to graduate from the program (82% vs. 49%). Only GRE scores were different for International vs. Domestic students when examining doctoral students. For M.A. students, GRE-V was positively correlated with GGPA and GRE-Q was positively related to earning the M.A. degree. Regression analyses controlling for several factors, including domestic student status, found only undergraduate GPA to predict GGPA and graduation for M.A. students. GGPA and GRE-V predicted graduation rates for M.A. students when GGPA was included in logistic regression analysis. GRE, GPA, and GGPA failed to predict Ph.D. success when examined together.

Keywords:

• GRE
• Communication
• Validity
• Graduate
• Success

Notes

Thomas Hugh Feeley (Ph.D., University at Buffalo, 1996) is Research Associate Professor in Department of Communication at University at Buffalo where Vivian M. Williams (Ph.D., University at Buffalo, 2004) is Assistant to the Chair and Timothy J. Wise (B.A., University at Buffalo, 2003) is an M.A. student.

The authors would like to thank Dr. Kathryn Olson for her comments on an earlier version of this manuscript.

The authors would like to thank Dr. Kathryn Olson for her comments on an earlier version of this manuscript.

When comparing GRE scores for international and domestic student status in three cases (GRE-Q for M.A. & Ph.D., GRE-V for M.A.) were the variances unequal (using Levene's test in ANOVA), thus violating an assumption of ANOVA. We tried several methods to transform the data to normalize the variances but met with little success (e.g., Smith, 1976). In all cases there was a need to eliminate outliers to standardize the scores but the gains from doing this were offset, in our opinion, by including all data in analyses. As our goal was to study all students and to generalize to future application cohorts, we felt it necessary to include all data. Thus all data were not transformed and all means are actual not transformed. It is our estimation that transforming the data would serve to actually raise the effect sizes between domestic/international status and study factors—this is important and interesting certainly but the mitigating role of international student status was not focal point of study and all statistical tests were computed controlling for this factor.

Additional information

Notes on contributors

Timothy J. Wise

Thomas Hugh Feeley (Ph.D., University at Buffalo, 1996) is Research Associate Professor in Department of Communication at University at Buffalo where Vivian M. Williams (Ph.D., University at Buffalo, 2004) is Assistant to the Chair and Timothy J. Wise (B.A., University at Buffalo, 2003) is an M.A. student. The authors would like to thank Dr. Kathryn Olson for her comments on an earlier version of this manuscript.