Date created: | Last Updated:

Category: Project

Description: Systematic Reviews and meta-analyses have been on top of the hierarchy for evidence synthesis in medicine. Statistical methods for analysis of meta-analysis have often been through the use of a classical framework. Nonetheless here we showcase an alternative Bayesian model averaging approach through a large scale re-analysis of continuous outcome meta-analyses from the Cochrane Database of Systematic Reviews. Moreover, different vague and informed priors for the effect size parameter δ and the between-study variance parameter τ are proposed and tested by means of an hold-out validation strategy. In addition, this proposal presents a design for conducting a sensitivity analysis for comparing the Bayesian model averaging methods against traditional methods and measures. Overall, it is proposed to entertain the same methods and models on a validation set to showcase the potential use of a Bayesian model averaging approach.