Analysis of the probability distribution method for spectrum decomposition

Abstract

The probability distribution method provides a new way for spectrum deconvolution, that is especially applicable in the case of poor counting statistics. Here it has been compared with the conventional least-squares curve-fitting method. A large number of spectrum regions of a few channels were generated to simulate a spectrum containing a low intensity peak on low background, and were evaluated by both methods, to establish which method provides an unbiased estimate of the peak area in the cases examined, and on what conditions.