Practical aspects of Bayesian multivariate meta-analysis

Abstract

Multivariate meta-analysis is a mostly used approach when multivariate results of several studies are pooled together. The multivariate model of random effects provides a tool to perform the multivariate meta-analysis in practice. In this paper, we discuss Bayesian inference procedures derived for the multivariate model of random effects when the model parameters are endowed with two non-informative priors: the Berger-Bernardo reference prior and the Jeffreys prior. Moreover, two Metropolis-Hastings algorithms are presented, and their convergence properties are analysed via simulations.