Effects of Restriction of Range on Predictive Validity for the Graduate Record Examination

Abstract

Several discussions of the effects of restriction of range on predictive validicy have been presented (1. 3). Generally, the presence of restriction of range on a predictor test will produce lower than expected predictive validity for the performance measure. T o estimate more accurately true predictive validity when restriction of range is present, several correction formulas have been presented (1 , 2 ) . A common area in which restriction of range becomes of interest is in psychological measurement, such as the relationship between Graduate Record Examination (GRE) scores and subsequent graduate grade point average. Although restriction effects have been widely discussed, relatively few studies have determined the magnitude of the effects. This study assessed the effects of the extent of restriction of range in a common measurement situation, the effects o n the predictive validity of the GRE. Graduate grade point averages and GRE scores were obtained for 8 5 students who were accepted for graduate smdy in a Master's level program in psychology at a large midwestern university. GRE scores were also obtained for an additional 72 s u b jects who were denied admission. Using a formula to correct for restriction which was presented by Givner and Hynes ( I ) , both corrected and uncorrected predictive validities were computed. The uncorrected predictive validity of the GRE was .15, and che predictive validicy corrected for restriction effects increased slightly to .17; neither correlation is significant. There is no evidence that improvement in a correlation corrected for restriction is dependent upon the magnitude of the original predictive validity. The relatively small increase in predictive validity in this study suggests that, in many actual measurement siruations, effects of restrictions of range may be slight.