Two methods of estimating the predictive validity of college admissions tests are presented. The first is a multivariate extension of a method presented by Givner and Hynes. The second allows an estimate of the predictive validity of one variable assuming that the other variable is a nationally representative sample. (Author/MLW)

A study of Black and White 1974 freshmen explored computing graduation predictions for admissions use. Sex, race, residency, major, high school rank, high school size, Scholastic Aptitude Test scores, and predicted grade point average were used as predictors. Correct predictions occurred for 64.5 percent of Blacks, 57.65 percent of Whites. (MSE)