A comment on the general inapplicability of Ghiselli's moderator system for two predictors

Abstract

A. Conger (see record 1970-04335-001 ), in an attempt to formalize the 2-predictor moderated regression system proposed by Ghiselli, concluded that the correlation between the difference scores should be near 0 for successful application of Ghiselli's system. Theoretical results are presented which indicate that this correlation coefficient would usually be much closer to unity. A real data example is presented that illustrates this. In view of this result and other considerations, the applied investigator is advised to employ alternative models for moderated regression. (PsycInfo Database Record (c) 2025 APA, all rights reserved)