The Meta-Analysis of Bivariate and Partial Effects Using a Power Prior Distribution

• In Event: Marriott Roundtable Session 65
  In Roundtable Session: New Methods for the Meta-Analysis of Regression Coefficients

Abstract

Several methods have been proposed for the meta-analysis of regression studies. One of these approaches synthesizes partial effects computed from regression model output. When a collection of studies has a mixture of bivariate effects and partial effects, typical practice is to perform separate meta-analyses based on the type of effect size. This paper proposes a method which combines bivariate and partial effects in a single Bayesian meta-analysis using a power prior distribution approach. We discuss the power prior distribution methodology using two sets of scenarios: bivariate ideal model and multivariate ideal model. Several power prior weighting schemes are proposed and examined.

Authors

• Christopher Glen Thompson, Texas A&M University - College Station

• Betsy J. Becker, Florida State University