A Robust Hand-Calculated Bayesian Analysis

Abstract

Robustness is a term attached to methods that are insensitive to assumptions extraneous to what is being studied. For example, in continuous distributions the shape of the distribution of sample medians is insensitive to the shape of the tails of the distribution from which the sample is drawn, and therefore medians are robust against changes in tails. Our main study depends upon distributional assumptions, such as the Poisson or negative binomial, and though we have studied their appropriateness, still it would be well to have a method that is less sensitive to distributional shape. Of course, we cannot expect from the robust study the strength of discrimination of the main study.