Online Program

Predictive Modeling for Over-Dispersed Proportion Data Using Some Completing Proportion Models (304080)

Keywords: Beta-binomial, correlated binomial model, double binomial model, extra-dispersion, toxicological data.

Proportion data occurring in many applied fields exhibit extra-variation predicted by a simple binomial model. For modeling extra-dispersed proportions, many authors have introduced several alternative extra-dispersed proportion models. With real-life data, a practical problem is deciding how to select one out of a wide variety of candidate models. In this paper, we aim to solve this problem in terms of real-life data occurring in a toxicological study. We discuss the model selection issues using a variety of standard model selection approaches. Moreover, a parametric bootstrap approach of model evaluation using a Mahalanobis squared distance proposed by Allcroft and Glasbey (Statistical Modelling, 2003) is applied.