A note on a Bayesian procedure for meta-analysis of rare data

Abstract

The propose of this paper is to develop a Bayesian procedure that adequately account for studies with zero observations in meta-analysis and then we focus the problem in the context of the Bayesian selection models. Also, attention is focused to the link distribution between effectiveness in each study/center and the meta-effectiveness. We present an objective Bayesian method where all quantities of inter-est jointly with a Bayesian test for equality between treatments are also obtained. A couple of examples with is developed in depth using the pro-posed Bayesian meta-analysis for the binomial model. Basically, we obtain a Bayesian model for meta-analysis for sparse binomial data without consid-ering transformations and/or corrections in variable/parameters. In respect to the examples considered, we do not find a relevant difference between treatments.