A Table of Correction Factors for Estimating Point-Biserial Correlations with Equal Category Proportions

Abstract

A table of correction factors is presented for removing the downward bias in a point-biserial correlation resulting from an unequal number of observations in each category of the dichotomous variable. The procedure involves calculating the observed point-biserial correlation, transforming it to the biserial correlation, and then recomputing the point-biserial correlation with a 50–50 split. The latter two steps can be accomplished by use of the table presented herein. When conceptually appropriate, this procedure is recommended for obtaining more accurate estimates of population correlations, a priori and post hoc power analysis, and for removing the effect of disparate splits prior to aggregation in meta-analysis.