Averaging Untransformed Variance Ratios Can Be Misleading: A Comment on Feingold

Abstract

Studies involving gender differences in cognitive ability have almost exclusively examined mean differences. In a recent article, Feingold (1992a) addressed this deficiency in the literature with an examination of variability differences between males and females on several cognitive test batteries. His analyses consisted of calculations of variance ratios: dividing male variances by female variances. Averaging these variance ratios, however, can yield misleading conclusions. The creation of log-transformed variance ratios is a simple solution to this problem.