J. de Uña Álvarez, M. D. Jiménez Gamero
The Poisson distribution is often used to perform statistical inference from count data. Needless to say, testing for the Poisson model is critical to validate the inference in such setting. Several goodness-of-fit tests for the Poisson distribution exist; they work for a single sample and an increasing sample size. However, in some applications the Poisson model is assumed for a large number k of populations, from which samples with a relatively small size n are available. In this work the simultaneous testing for the Poisson distribution along k populations is considered. A test statistic that combines the Baringhaus-Henze scores computed from the k samples is proposed. The null distribution of the test as k goes to infinity is obtained. The consistency of the test is established. The results cover the independent case and also the case of weak dependence along the k samples. The practical performance of the test is investigated through simulations and real data analyses.
Keywords: Count data, Goodness-of-fit, Multiple testing, NGS experiments
Scheduled
GT09.NOPAR4 Invited session.Nonparametric tests
November 9, 2023 3:30 PM
CC1: Audience