ABSTRACT
This chapter describes the Bayesian approach to meta-analysis, where inferences are based on a posterior distribution that describes knowledge about model parameters, conditional on the data and prior or external information. Statements about the probability of scientific hypotheses follow directly from the posterior distribution. Topics discussed include choosing the prior distribution, summarizing the posterior distribution by point and interval estimates, computing the posterior with Markov chain Monte Carlo, making predictive inferences, and assessing model fit and assumptions. Bayesian methods for common- and random-effects models using one- and two-stage methods are applied to examples for continuous and discrete outcomes fit with R and JAGS software. Advantages and disadvantages of the Bayesian approach are presented and contrasted with the frequentist approach.
