Meta-Analysis: A Structural Equation Modeling Approach

Chapter 2Brief review of structural equation modeling

This chapter reviews selected topics in structural equation modeling (SEM) that are relevant to the SEM-based meta-analysis. It provides a quick introduction to SEM for those who are less familiar with the techniques. This chapter begins by introducing three different model specifications—path diagrams, equations, and matrix specification. It then introduces popular structural equation models such as path analysis, confirmatory factor analytic (CFA) models, SEMs, latent growth models, and multiple-group analysis. How to obtain parameter estimates, standard errors (SEs), confidence intervals (CIs), test statistics, and various goodness-of-fit indices are introduced. Finally, we introduce phantom variables, definition variables, and full information maximum likelihood (FIML). These concepts are the keys to formulating meta-analytic models as structural equation models.

2.1 Introduction

SEM, also known as covariance structure analysis and correlation structure analysis, is a generic term for many related statistical techniques. Many popular multivariate techniques, such as correlation analysis, regression analysis, analysis of variance (ANOVA), multivariate analysis of variance (MANOVA), factor analysis, and item response theory, can be considered as special models of SEM. Generally speaking, SEM is a statistical technique to model the first and the second moments of the data when the data are multivariate normal. The first ...