Department of Biostatistics Seminar/Worki Progress Series

What Does a Bayesian Owe a Frequentist?

Frank E. Harrell, Jr., Ph.D

Professor and Chairman, Department of Biostatistics

Vanderbilt University School of Medicine

This is a work in progress in preparation for a national presentation at the American Course on Drug Development and Regulatory Sciences sponsored by UCSF and FDA, in a session entitled "Opportunities to Advance the Use of Bayesian Methods for Drug Development and Research". Your feedback and ideas will be highly valued.

The talk will start with a brief overview of the problems with P-values, especially related to "the multiplicity mess." I'll discuss problems caused by the frequentist's need to reverse the flow of time and information and to constantly account for the sample space. Then I'll show simple examples of continuous monitoring of clinical trials where irrelevance of type I error and example of posterior probability paths are given. I'll conclude with an appraisal of what a Bayesian owes a frequentist statistician or regulator, and a brief statement of what a frequentist owes a Bayesian. An underlying theme is that the FDA CDRH guidance document on the use of Bayesian methods in device trials, which attempts to incorporate a hybrid Bayesian/frequentist approach, is holding back progress and creating unnecessary work for statisticians. Scott Berry will speak after me, so the last topic is bound to be quite controversial in view of the large amount of time he spends simulating frequentist operating characteristics of Bayesian clinical trials.

Topic revision: r1 - 01 Apr 2016, AshleeBartley
 

This site is powered by FoswikiCopyright © 2013-2017 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback