Dados do Trabalho

Título

IMPROVED BAYES ESTIMATORS FOR THE WILSON-HILFERTY DISTRIBUTION

Resumo

In this work, we revisit the Wilson-Hilferty distribution. The proposed model is useful to describe devices with high chance of failure in the very early stages, but the hazard function levels off and eventually increases as the system get old. The maximum likelihood estimators are explored under complete and censoring. Additionally, we present a Bayesian reference analysis for the generalized gamma distribution by using a reference prior, which has important properties such as one-to-one invariance under reparametrization, consistent marginalization, consistent sampling and leads to a proper posterior density. A simulation study compares the performance of the estimators with a clear advantage for the Bayesian approach.

Palavras-chave

Wilson-Hilferty distribution; Maximum likelihood estimators; Objective Prior, Reference Prior.