A Bayesian approach to exploratory analysis in clinical trials
Exploratory analysis consumes much of the analysis effort in clinical trials, yet it is difficult to interpret and cannot be generalized to the population from which the research sample was obtained. However, due to the random selection of endpoints, the likelihood principle conditions have been violated in exploratory analysis. In this case, the likelihood function and Bayesian approach has to been revised to incorporate the randomness of endpoint selection. The primary objective of our study is to demonstrate the failure of maximum likelihood estimates in the exploratory analysis paradigm and to develop alternative estimates using a Bayes approach. Two cases are provides. 1) the beta prior and binomial likelihood function and 2) gamma prior and Poisson likelihood function. Weighted loss functions will be applied to more realistically reflect the peril faced by investigators when disease burden is underestimated in the exploratory setting.
Han, Weilu, "A Bayesian approach to exploratory analysis in clinical trials" (2014). Texas Medical Center Dissertations (via ProQuest). AAI1599376.