An Evaluation of Multivariate Methods for Meta-Analytic Structural Equation Modeling

• In Event: Fancy Structural Equation Modeling and Hierarchical Linear Modeling

Abstract

While two-stage structural equation modeling (TSSEM) is widely used for meta-analytic structural equation modeling (MASEM), alternative multivariate meta-analytic methods could also be used for MASEM. The present study evaluated the performance of robust variance estimation (RVE), three-level multivariate meta-analysis with RVE (3MLM-RVE), and TSSEM, assuming random-effects model, in terms of parameter and variance recovery rates in pooled correlation estimates and the structural parameters under realistic conditions. Findings from this study showed that RVE and TSSEM provided unbiased pooled correlations, whereas pooled correlations from 3MLM-RVE were positively biased. Further, TSSEM and RVE provided unbiased structural parameter estimates. RVE provided unbiased standard errors (SEs) with moderate to large number of studies, whereas TSSEM provided unbiased SEs with a small number of studies.

Authors

• Kejin Lee, University of Illinois at Chicago

• Tasha Beretvas, The University of Texas at Austin