STATISTICAL METHODS APPLIED TO PSD DATA : CHARACTERIZATION ON THE DETECTOR RESPONSE

Abstract

Several procedures are considered for testing numerically the performances of some two-dimensional position sensitive gas detectors (PSD). The response of a PSD to a homogeneous flux of scattered neutrons is modelled by a two-dimensional Poisson process and specific statistical tests and techniques are developed in order to determine the PSD's characteristics. Two approaches are used : the first one employs a non-parametric probability scheme for recognizing the inhomogeneity in the experimental data. - The second one extends the classical two-way analysis of variance to a more general Poisson scheme using the maximum likelihood principle. The temporal stability of PSD data is also tested by means of a new two-sample comparison test on Poisson data. Numerical examples from simulated data are presented and the performances of the statistical algorithms are discussed.