A group-specific prior distribution for effect-size heterogeneity in meta-analysis.

Area of Science:

• Social Sciences, Psychology, Education
• Statistical Modeling
• Meta-Analysis

Background:

• Bayesian meta-analysis methods are advancing, but modeling study group differences remains limited.
• Existing Bayesian approaches often use non-informative priors, which may not be optimal.
• There is a growing interest in utilizing more informative prior distributions.

Purpose of the Study:

• To propose a novel group-specific, weakly informative prior distribution for the between-studies standard deviation parameter in meta-analysis.
• To enhance Bayesian meta-analysis for psychology, education, and social sciences by modeling subgroup differences.
• To incorporate frequentist estimates into a folded noncentral t distribution for prior modeling.

Main Methods:

• Developed a group-specific weakly informative prior distribution for the between-studies standard deviation.
• Utilized a folded noncentral t distribution, incorporating frequentist estimates as the noncentrality parameter.
• Applied and compared the proposed prior to non-informative priors in two meta-analysis examples from psychological interventions.

Main Results:

• Demonstrated the application of the proposed folded noncentral t prior distribution in meta-analysis.
• Provided empirical comparisons against several non-informative prior distributions.
• Highlighted the utility of the new prior for modeling heterogeneity across study subgroups.

Conclusions:

• The proposed group-specific prior distribution offers a valuable advancement for Bayesian meta-analysis in social sciences.
• This method improves the modeling of between-study heterogeneity when studies can be meaningfully grouped.
• Further research is encouraged to expand Bayesian meta-analysis techniques in psychology and related fields.