Search

The ETS has recently released new estimates of validities of the GRE for predicting cumulative graduate GPA. They average in the middle thirties - twice as high as those previously reported by a number of independent investigators. It is shown in the first part of this paper that this unexpected finding can be traced to a flawed methodology that tends to inflate multiple correlation estimates, especially those of populations values near zero. Secondly, the issue of upward corrections of validity estimates for restriction of range is taken up. It is shown that they depend on assumptions that are rarely met by the data. Finally, it is argued more generally that conventional test theory, which is couched in terms of correlations and variances, is not only unnecessarily abstract but, more importantly, incomplete, since the practical utility of a test does not only depend on its validity, but also on base-rates and admission quotas. A more direct and conclusive method for gauging the utility of a test involves misclassification rates, and entirely dispenses with questionable assumptions and post-hoc "corrections". On applying this approach to the GRE, it emerges (1) that the GRE discriminates against ethnic and economic minorities, and (2) that it often produces more erroneous decisions than a purely random admissions policy would.

Scargle (2000) has discussed Rosenthal and Rubin's (1978) "fail-safe number" (FSN) method for estimating the number of unpublished studies in meta-analysis. He concluded that this FSN cannot possibly be correct because a central assumption the authors used conflicts with the very definition of publication bias. While this point has been made by others before (Elsahoff, 1978; Darlington, 1980; Thomas, 1985, Iyengar & Greenhouse, 1988), Scargle showed, by way of a simple 2-parameter model, how far off Rosenthal & Rubin' s estimate can be in practice. However, his results relied on the assumption that the decision variable is normally distributed with zero mean. In this case the ratio of unpublished to published papers is large only in a tiny region of the parameter plane. Building on these results, we now show that (1) Replacing densities with probability masses greatly simplifies Scargle's derivations and permits an explicit statement of the relation between the probability alpha of Type I errors and the step-size beta; (2) This result does not require any distribution assumptions; (3) The distinction between 1-sided and 2-sided rejection regions becomes immaterial; (4) This distribution-free approach leads to an immediate generalization to partitions involving more than two intervals, and thus covers more general selection functions., Comment: 18 pages, 3 figures