The Choice of Prior Distribution for a Covariance Matrix in Multivariate Meta-Analysis: A Simulation Study

• Activity Number: 53
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Madhu M. Mazumdar and Sandra M. Hurtado Rúa *+ and Robert Strawderman
• Companies: Mount Sinai School of Medicine and Weill Medical College of Cornell University and University of Rochester
• Keywords: Bayesian analysis ; priors for covariance matrices ; simulation

Abstract

Bayesian meta-analysis is an increasingly important component of clinical research, with multivariate meta-analysis a particularly promising tool for studies with multiple endpoints. Model assumptions, including the choice of priors, are crucial aspects of multivariate Bayesian meta-analysis (MBMA) models. In a given a model, two different prior distributions can lead to different inferences about a particular parameter. A simulation study was performed in which the impact of families of prior distributions of the covariance matrix of a multivariate normal random effects MBMA model was analyzed. We demonstrate the need for sensitivity analysis and propose guidelines regarding the use of prior distributions. Inferences about effect sizes were not particularly sensitive to prior choice, but the related covariance estimates were. A family of prior distributions with small relative biases, tight mean squared errors and closed to optimal coverage for the effect size estimates was identi?ed. The MBMA models proposed here are illustrated in a small meta-analysis example from the periodontal field.