5.3 Bayesian Model Averaging Continual Reassessment Method

In the CRM, the prespecified skeleton open , p sub 1 , comma dot dot dot comma , p sub j , close is usually arbitrary and subjective. As a result, the performance of the CRM may be compromised due to model misspecifications, which is a common drawback of parametric models. To enhance the robustness of the CRM, we can construct multiple CRM models in parallel, such that each model is equipped with a different skeleton representing uncertainty in the parametric modeling structure [3]. If there is no preference on different modeling structures, then we assign a discrete uniform prior probability to each CRM model. Through the Bayesian model averaging ...

Get Modern Approaches to Clinical Trials Using SAS: Classical, Adaptive, and Bayesian Methods now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.