Testing Moderator Variable Hypotheses Meta-Analytically

Abstract

We propose and illustrate a three-step procedure for testing moderator variable hypotheses meta-analytically. The procedure is based on Hedges and Olkin's (1985) meta-analytic approach, yet it incorporates study-level corrections for methodological and statistical artifacts that are typically advocated and used within psychometric approaches to meta-analysis (e.g., Hunter & Schmidt, 1990). The three- step procedure entails: (a) correcting study-level effect size estimates for across-study variability due to methodological and statistical arti facts, (b) testing the overall homogeneity of study-level effect size esti mates after the artifactual sources of variance have been removed, and (c) testing the effects of hypothesized moderator variables.