A Metric for Integrating Theoretically Related but Statistically Uncorrelated Constructs

Abstract

When considering psychological constructs in a pairwise fashion, a 2x2 category system, based on median splits, is often adopted for analytic convenience. If theoretical focus is on an individual's integration of the two constructs, then categorical procedures are subject to information loss and unreliability in high density regions of the data. A continuous metric is proposed which ameliorates these problems and is suitable for correlational analyses. The construct of psychological androgyny is used for illustrative purposes.