Meta-Analysis: A Structural Equation Modeling Approach

Chapter 9Conducting meta-analysis with Mplus

Most users of structural equation modeling (SEM) are familiar with at least one popular SEM package, such as Mplus, LISREL, or EQS. This chapter illustrates how to analyze the meta-analytic models introduced in previous chapters with Mplus. We show how Mplus can be used to conduct univariate, multivariate, and three-level meta-analyses using a transformed variables approach. Although we use Mplus in this chapter, the proposed transformed variables approach can also be applied to other SEM packages to conduct some of the SEM-based meta-analysis.

9.1 Introduction

Mplus (Muthén and Muthén, 2012) provides a unified framework to conduct data analyses using a latent variable modeling approach. It combines SEM, multilevel models, mixture modeling, survival analysis, latent class models, categorical variables, missing data analysis, item response theory models, robust test statistics, and Bayesian analysis into a single statistical modeling framework. The main strength of a unified framework to conduct data analysis is that some of these techniques can be combined together to address the research questions. For example, researchers may handle the missing data with maximum likelihood (ML) estimation method and the nonnormal data with robust statistics in the same analysis. It will be beneficial to methodologists and applied researchers if meta-analytic models can also be integrated as part of the SEM framework. The combination of meta-analysis ...