CORRECTING POINT-BISERIAL TURNOVER CORRELATIONS FOR COMPARATIVE-ANALYSIS

Several methods have been proposed for correcting point-biserial turnover correlations for variations across studies in turnover base rates, which artifactually constrain the size of such correlations. Williams (1990) critically evaluated these proposed methods and also questioned the appropriateness of such corrections when turnover base rates are "related to nonartifactual differences in conditions across studies" (p. 735). We point out additional problems with the proposed methods for correcting point-biserial correlations and with Williams' critique, and we suggest an alternative method for dealing with these problems.