A Comparison of a Fixed-Effects Model and Random-Effects Model for a Multivariate Structural Equation Modeling–Based Meta-Analysis

• In Event: 42.034 - Estimation and Model Advances for Causal Inference and Treatment Effects

Abstract

This study is to compare the FE and RE model for synthesizing multivariate effect sizes in the framework of structural equation modeling. Monte Carlo simulations were conducted to examine that in the heterogeneous case, what are the performances of mean effect size, standard error, Type I error rate for the two models under different data conditions. Results indicate that there is no difference of the value of mean effect size between FE model and RE model. But the RE model demonstrated a higher accuracy for the mean standard error. Compared to FE model, RE model can robustly control the type I error. Increasing primary study sample sizes and between-studies variances resulted in increasingly inflated Type I error for FE model.

Authors

• Zhihui CAI, University of Macau

• Xitao Fan, The Chinese University of Hong Kong (Shenzhen)