Uncertainty Analysis of Analytical Results When Errors are Not Normally Distributed*

This paper presents a method for calculating a priori uncertainty intervals for plutonium interpretations by fission track analysis (FTA) of human urine samples. The derivations of the method are adaptable to estimating uncertainty intervals of analysis of single samples, based on calibration data obtained from multiple samples spiked over a range of concentration. The data used in last year’s BAER conference talk on the population excretion of plutonium had approximately normally-distributed errors in tracks observed at given spike levels, with almost constant variance in tracks vs. plutonium spike level. This allowed fitting of the pre-September 1998 calibration line by a simple linear regression. Methods are presented here for estimating a priori uncertainty intervals about interpretations using post-September 1998 data, which had non-normal error distributions and variance increasing with spike level.