Use of empirical Bayes methods in the study of the validity of academic predictors of graduate school performance