Methods for Meta-Analytic Structural Equation Modeling

Abstract

The process of performing meta-analytic structural equation modeling (MASEM) consists of two stages. First, correlation coefficients that have been gathered from studies have to be combined to obtain a pooled correlation matrix of the variables of interest. Next, a structural equation model can be fitted on this pooled matrix. Several methods are proposed to pool correlation coefficients. In this chapter, the univariate approach, the generalized least squares (GLS) approach, and the Two Stage SEM approach are introduced. The univariate approaches do not take into account that the correlation coefficients may be correlated within studies. The GLS approach has the limitation that the Stage 2 model has to be a regression model. Of the available approaches, the Two Stage SEM approach is favoured for its flexibility and good statistical performance in comparison with the other approaches.