Title: Synthesizing Contextual Effects: A Comparison of the Two-Stage Meta-Analytic and One-Stage Multilevel Modeling Approaches in the presence of Individual Participant Data
Abstract: Group composition effects are a central focus of social science research. Nevertheless, the generalizability of these effects across populations and settings remains unclear. Multilevel Models (MLM) and two-stage individual participant data (IPD) meta-analysis are two commonly used approaches to estimate and synthesize generalizable contextual effects. However, MLM with random slopes are computationally slow, imprecise, and frequently lead to convergence problems. Moreover, empirical evidence regarding the performance and accuracy of the two-stage IPD meta-analysis approach—a viable alternative to MLM—is still lacking. In this study, we compare and evaluate the statistical performance of the two-stage meta-analytic and one-stage multilevel modeling approaches for synthesizing contextual effects in social science research using individual participant data