ABSTRACT

This chapter discusses the specification of the prior standard deviation σβ for the model parameter β in a Bayesian CRM. Section 9.2 introduces the notion of least informative prior in the context of dose finding, and summarizes the guidelines on the specification of prior distribution in terms of its variability. We consider two calibration approaches: Section 9.3 presents an approach informed by the use of the least informative prior for a given sensitivity δ , and Section 9.4 optimizes the least informative model with respect to δ . Section 9.5 compares these approaches in the context of the bortezomib trial.