A Bayesian analysis of zero and one inflated distributions

Milton Yoshio SAITO[1]


§       abstract: In this paper, we study  zero and one inflated discrete distributions under a Bayesian point of view. We introduce two simple latent variables which simplify the expression of the likelihood function. An ilustrative example using the Poisson model and a data set of Swedish dental insurance system are presented. The MCMC algorithm is adapted to a Bayesian solution to the problem and later samples are easily obtained..

§       keywords: Bayesian Inference; zero and one excess; mixture.

[1] Departamento de Estatística, Universidade Federal de São Carlos - UFSCar, Caixa Postal 676, CEP 13565-905, São Carlos, SP, Brasil. E-mail: vjosemar@power.ufscar.br