Priors for Between-Study Heterogeneity to Prevent Overconfidence in Bayesian Bias-Adjustment Meta-Analysis Models

• In Event: Advances in Meta-Analytic and Systematic Review Approaches

Abstract

This research enhances Bayesian bias-adjustment models in meta-analysis by incorporating risk of bias (RoB) assessments and addressing estimate overconfidence. It introduces a prior distribution for between-study heterogeneity to reduce estimate distortion and prevent narrow credible intervals (CrIs). Real data analysis shows that a Gamma prior with specific parameters (α = 2, λ ≥ 10) yields smaller overall effect size estimates and wider CrIs, indicating increased uncertainty and larger between-study heterogeneity. These findings align with theoretical expectations and underscore the importance of careful prior selection. The study preserves RoB assessment information and offers a practical solution to potential model paradoxes, improving the robustness and reliability of evidence-based decision-making in bias-prone educational research.

Authors

• Juyoung Jung, University of Iowa

• Ariel M. Aloe, University of Iowa