Evaluating a Bayesian Approach for Estimating Moderator Effects in Parameter-Based Meta-Analytic Structural Equation Modeling

Meta-analytic structural equation modeling (MASEM) enables meta-analytic investigations of multivariate models. A common research objective in meta-analyses is identifying study-level moderators that explain heterogeneity across primary studies. Several MASEM approaches have been extended to include moderators, however research evaluating and comparing these approaches remains scarce. The present study discusses several parameter-based moderated MASEM approaches covering one-stage and two-stage as well as fixed and random effects approaches. We implement these different MASEM approaches using a Bayesian modeling framework and evaluate them through two simulation studies. We compare bias, efficiency, coverage rates, and statistical power of moderator effects across varying number of primary studies, sample sizes, random effect structures, and structural models. Results imply that different parameter-based MASEM approaches can provide unbiased estimates and appropriate coverage of study-level moderation effects. Exceptions and implications are discussed.