Setting confidence belts

Abstract

We propose using a Bayes procedure with uniform improper prior to determine\ncredible belts for the mean of a Poisson distribution in the presence of\nbackground and for the continuous problem of measuring a non-negative quantity\n$\\theta$ with a normally distributed measurement error. Within the Bayesian\nframework, these belts are optimal. The credible limits are then examined from\na frequentist point of view and found to have good frequentist and conditional\nfrequentist properties.\n