Inferência no modelo potência normal bimodal assimétrico

Abstract

This work is a further study of the model recently proposed by Bolfarine et al. (2013), the Bimodal Asymmetric Power Normal model (PNBA). After a review of the basic properties of this model, the following points were approached: development of an algorithm for PNBA random number generation, comparison of maximum likelihood and method of moments in small samples and a first investigation of a resampling method (bootstrap) to compare two PNBA populations. Numerical difficulties related to maximum likelihood estimation due to local maxima in the likelihood function were discussed. Finally, an application to real data illustrated the use of the model.