Comparison of Three Meta-Analytic Procedures for Estimating Moderating Effects of Categorical Variables

Abstract

The authors conducted Monte Carlo simulations to compare the Hedges and Olkin, the Hunter and Schmidt, and a refinement of the Aguinis and Pierce meta-analytic approaches for estimating moderating effects of categorical variables. The simulation examined binary moderator variables (e.g., gender—male, female; ethnicity—majority, minority). The authors compared the three meta-analytic methods in terms of their point estimation accuracy and Type I and Type II error rates. Results provide guidelines to help researchers choose among the three meta-analytic techniques based on theory (i.e., exploratory vs. confirmatory research) and research design considerations (i.e., degree of range restriction and measurement error).