Predictive validities: figures of merit or veils of deception - Statistics > ApplicationsReport as inadecuate




Predictive validities: figures of merit or veils of deception - Statistics > Applications - Download this document for free, or read online. Document in PDF available to download.

Abstract: The ETS has recently released new estimates of validities of the GRE forpredicting cumulative graduate GPA. They average in the middle thirties - twiceas 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 betraced to a flawed methodology that tends to inflate multiple correlationestimates, especially those of populations values near zero. Secondly, theissue of upward corrections of validity estimates for restriction of range istaken up. It is shown that they depend on assumptions that are rarely met bythe data. Finally, it is argued more generally that conventional test theory,which is couched in terms of correlations and variances, is not onlyunnecessarily abstract but, more importantly, incomplete, since the practicalutility of a test does not only depend on its validity, but also on base-ratesand admission quotas. A more direct and conclusive method for gauging theutility of a test involves misclassification rates, and entirely dispenses withquestionable assumptions and post-hoc -corrections-. On applying this approachto the GRE, it emerges 1 that the GRE discriminates against ethnic andeconomic minorities, and 2 that it often produces more erroneous decisionsthan a purely random admissions policy would.



Author: Peter H. Schonemann, Moritz Heene

Source: https://arxiv.org/







Related documents