Abstract:
|
The motivation is to examine the clinical applications of Bayseian theory in CMA and to develop methods of applying this theory to safety data. We developed and applied a hybrid approach, which combine the Bayesian hierarchical modeling and frequentist approach for meta-analysis. The question with regards to selecting likelihood function will be presented. The analysis is conducted using the example of CMA regarding risk management safety data. The hybrid analysis is derived by verifying the assumptions through a frequentist approach. The implications of this combined approach are discussed, including using Peto's method of Q heterogeneity statistics to determine exchangeability. This analysis will also address the process of determining the proper Bayesian prior and likelihood distributions. Finally, Bayesian methods are illustrated using graphical methods to examine extreme probabilities in safety data. The proper Bayesian distribution is determined through these safety graphs.
|