Accommodating Uncertainty in Prior Distributions
Los Alamos National Lab. (LANL), Los Alamos, NM (United States), Richard Picard, USDOE Assistant Secretary for Human Resources and Administration, Scott Vander Wiel · 2017
A fundamental premise of Bayesian methodology is that a priori information is accurately summarized by a single, precisely de ned prior distribution. In many cases, especially involving informative priors, this premise is false, and the (mis)application of Bayes methods produces posterior quantities whose apparent precisions are highly misleading. We examine the implications of uncertainty in prior distributions, and present graphical methods for dealing with them.