Summarizing empirical information on between-study heterogeneity for Bayesian random-effects meta-analysis

Area of Science:

• Biostatistics
• Statistical Modeling

Background:

• Bayesian meta-analysis frequently necessitates prior probability specifications for between-study heterogeneity.
• This is particularly crucial in meta-analyses involving a limited number of studies.
• Existing methods for utilizing historical data to inform these priors are often inadequate.

Approach:

• The study extends the standard normal-normal hierarchical model for random-effects meta-analysis.
• It introduces a method to infer a heterogeneity prior directly from empirical data.
• Focuses on simple, applicable strategies for fitting parametric distributions to observed heterogeneity data.

Key Points:

• Demonstrates fitting a distribution to empirically observed heterogeneity data from multiple meta-analyses.
• Highlights the importance of selecting an appropriate parametric distribution family.
• Provides practical guidance for translating empirical findings into prior probability distributions.

Conclusions:

• The proposed extension offers a more robust method for specifying priors in Bayesian meta-analysis.
• This approach enhances the reliability of meta-analysis results, especially in data-scarce scenarios.
• Facilitates the informed use of historical data for setting heterogeneity priors.