The Statistics of Low-Level Counting Using the New Generation of Packard Liquid Scintillation Counters

Abstract

We consider the suitability of commonly used Poisson counting statistics applied to background count rates measured in the new generation of low-background Packard liquid scintillation spectrometers. We also investigate the stability of these systems over long time intervals. Undetected instability will result in an underestimation of the precision of any result (i.e., the calculated error will be too small), and, in the presence of a systematic source, could lead to inaccurate results. The work described here forms only a small part of a project to investigate the statistical criteria that should be applied to the performance of such counters. The procedures to be discussed here include the Poisson index of dispersion, x and s control charts and the MSSD test for detection of drift. These are illustrated on background count rates derived from the Packard 2260XL and 2000CA/LL.