S3RI Seminar - Dr Apostolos Batsidis (University of Ioannina) Seminar

Event details

Given a sample from an unknown count population and two possibly misspecified parametric models, which may be separate, overlapping or nested, the problem of model selection consists in testing if the two competing models are equally close to the true population, against the hypothesis that one model is closer than the other. In this talk, measuring the closeness of a model to the true population by means of a distance based on the probability generating functions, a test approach to the model selection is presented. The motivation for using the probability generating function (PGF) is that it is usually much simpler than the corresponding probability mass function, fully characterizes the distribution and possesses convenient features not shared by the characteristic or moment generating function. In this context, the unknown parameters are estimated based on a pgf-based divergence. As a prerequisite, some asymptotic properties of these estimators are studied. From these properties, consistent tests for model selection based on PGFs are given for separate, overlapping and nested models. The finite sample performance of the proposed methods is evaluated by simulations.