Quantifying and explaining heterogeneity in meta-analytic structural equation modeling: Methods and illustrations.

Abstract

As a method for developing and testing hypotheses, meta-analytic structural equation modeling, or MASEM, has drawn the interest of scholars. However, challenges remain in how we can model and explain meaningful heterogeneity in structural equation modeling (SEM) parameters. To address this issue, two novel methods have recently been proposed in the literature: Bayesian MASEM (BMASEM) and one-stage MASEM (OSMASEM). How the two methods can be applied to address actual psychological research questions involving heterogeneity is a topic of debate and confusion. In this study, we describe and compare the two methods using two illustrations on the mediating mechanism of mindfulness-based intervention and the factor structure of Rosenberg Self-Esteem Scale. In the illustrations, both methods were used to test the moderating effect of a covariate, to build a prediction equation for effect sizes in specific populations, and to evaluate the equivalence of standardized factor loadings of a scale. The study ends with a discussion of practical issues that may arise when applying BMASEM and OSMASEM.