Meta-analysis of Monte Carlo simulations examining class enumeration accuracy with mixture models.

Abstract

This article walks through steps to conduct a meta-analysis of Monte Carlo simulation studies. The selected Monte Carlo simulation studies focused on mixture modeling, which is becoming increasingly popular in the social and behavioral sciences. We provide details for the following steps in a meta-analysis: (a) formulating a research question; (b) identifying the relevant literature; (c) screening of the literature; (d) extracting data; (e) analyzing the data; and (f) interpreting and discussing the findings. Our goal was to investigate which simulation design factors (moderators) impact class enumeration accuracy in mixture modeling analyses. We analyzed the meta-analytic data using a generalized linear mixed model with a multilevel structure and examined the impact of the design moderators on the outcome of interest with a meta-regression model. For instance, the Bayesian information criterion was found to perform more accurately in conditions with larger sample sizes whereas entropy was found to perform more accurately with smaller sample sizes. It is hoped that this article can serve as a guide for others to follow in order to quantitatively synthesize results from Monte Carlo simulation studies. In turn, the findings from meta-analyzing Monte Carlo simulation studies can provide more details about factors that influence outcomes of interest as well as help methodologists when planning Monte Carlo simulation studies. (PsycInfo Database Record (c) 2024 APA, all rights reserved).