Modelo de regressão com erros normais assimétricos: uma abordagem bayesiana

The statistical analysis of the continuous data set has been developed in most cases for normal models, specially, in the linear and nom linear models context. In the simple linear regression models, even accepting as reasonable the normal error assumption, usually it can take misleading inferences...

Descripción completa

Detalles Bibliográficos
Autor: Freitas, Luiz Antonio de
Tipo de recurso: tesis de maestría
Estado:Versión publicada
Fecha de publicación:2005
País:Brasil
Institución:Universidade Federal de São Carlos (UFSCAR)
Repositorio:Repositório Institucional da UFSCAR
Idioma:portugués
OAI Identifier:oai:repositorio.ufscar.br:20.500.14289/4592
Acceso en línea:https://repositorio.ufscar.br/handle/20.500.14289/4592
Access Level:acceso abierto
Palabra clave:Análise de regressão
Inferência bayesiana
CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA
Descripción
Sumario:The statistical analysis of the continuous data set has been developed in most cases for normal models, specially, in the linear and nom linear models context. In the simple linear regression models, even accepting as reasonable the normal error assumption, usually it can take misleading inferences for the parameter of interest. In the classical literature is supposed that the errors follow the normal distribution with zero mean and unknown variance. In this work, we intend to make the normality as-sumption flexible using a new class of asymmetric distributions. The first class considered was studied by Azzalini [3], which includes the normal distributions as a particular case. The second flexible class was introduced by Ma & Genton [22] and the third one is the extended normal distributions proposed by Arellano-Valle [2]. These two last ones had been enclosed as alternative proposals to the model of Azzalini [3]. With this purpose, inferences procedures based on Bayesian approaches, will be developed to estimate the parameters involved in the simple linear regression models with asymmetrical errors.