|
Abstract:
|
We present a Bayesian framework to assist regulators making decisions about new drugs. In this study, we build a Bayesian model to do a full benefit-risk balance analysis of clinical trial data. One major novelty is that we are able to model the whole joint distribution of the effects of the treatment. As a result we can include interdependencies between continuous and discrete variables, which was previously not considered in drug tests. We achieve this by introducing latent variables which may be viewed, also, as patient’s random effects. This application to Biopharmaceutics stems from the first piece of a series of research on modern Bayesian Inference. The aim is to create a unified methodological framework for Latent Variable Models suitable for factor analysis in categorical data. The resulting algorithms will provide a way to get posterior samples from the joint distributions of latent factors and parameters. It will also compute the (Bayesian) marginal likelihood, often referred to as evidence, to facilitate Bayesian model choice.
|
