A Mixture Random-Effect Modeling That Addresses Population Heterogeneity Among Studies in Meta-Analysis

• In Event: Informing Educational Policy and Practice Using Research Synthesis

Abstract

In meta-analysis, researchers often estimate heterogeneity between effect sizes across studies. If variability exists, study characteristics can be used to explain it. However, sometimes researchers fail to find relevant moderator variables. In order to check whether studies are grouped in underlying clusters, mixture models can be incorporated in the random-effects model. To get a first impression of the performance of mixture models in meta-analysis, effect sizes grouped in two or three clusters of studies were generated. Afterwards, a random-effects model and mixture models of 1-, 2- and 3-class were applied. Preliminary results show that mixture models successfully recover cluster means and their between-studies variance. This simulation study will be extended for covering more realistic conditions.

Authors

• Belén Fernández-Castilla, University of Leuven

• Jung Yeon Park, University of Leuven

• Wim Van den Noortgate, KU Leuven