Bayesian Methods for Interpreting Plutonium Urinalysis Data

Abstract

We discuss an internal dosimetry problem, where measurements of plutonium in urine are used to calculate radiation doses. We have developed an algorithm using the MAXENT method. The method gives reasonable results, however the role of the entropy prior distribution is to effectively fit the urine data using intakes occurring close in time to each measured urine result, which is unrealistic. A better approximation for the actual prior is the log-normal distribution; however, with the log-normal distribution another calculational approach must be used. Instead of calculating the most probable values, we turn to calculating expectation values directly from the posterior probability, which is feasible for a small number of intakes.