ABSTRACT
Structural equation modeling (SEM) is a statistical technique which can be used to test relationships between observed and latent variables. Meta-analysis, being based on a multilevel model, can also be formulated from a SEM perspective. This relationship can be exploited to “replicate” a random-effects meta-analysis in the form of a structural equation model. More importantly, however, it allows to perform meta-analyses which model more complex relationships between observed effect sizes.
This chapter provides a brief introduction into meta-analytic SEM, starting with a formulation of the random-effects model within a SEM framework. Furthermore, the approach is used to conduct a multivariate meta-analysis, as part of a hands-on exercise in R. Lastly, the two-stage meta-analytic SEM procedure is illustrated using a confirmatory factor analysis example.
