DiagnÃstico em modelos simÃtricos de regressÃo

AUTOR(ES)

Luis Hernando Vanegas Penagos

DATA DE PUBLICAÇÃO

2005

RESUMO

The statistical modelling using the assumption of normally distributed errors can be strongly influenced by extreme observations. This motivates the study on regression techniques that are roobust in presenceof this type pof observations. As an alternative, models in wich the error term belongs to the symmetrical class can be considered. This class of models is called symmetrical regression models. In these models, the tails of the normal distribution. In this work, we studied analytically and numerically the properties for two types of proposed residuals for symmetrical nonlinear regression models: deviance and quantal. Besides, we developed diagnostic methods on sysmetrical regression models with systematic nonlinear components using the global influence approach. Through this approach we obtain measures as the Cook distance, W-K statistic, one step approximation, likelihood displacement and some graphics methods

ASSUNTO(S)

distÃncia de cook, estatÃstica w-k symmetrical class estatistica w-k statistic symmetrical regression models systematic nonlinear components modelos simetricos de regressÃo componentes sistemÃticas nÃo-lineares cook distance classe simÃtrica

Documentos Relacionados

• Modelos lineares generalizados simÃtricos heterocedÃsticos
• CorreÃÃo de Bartlett nos modelos nÃo-lineares simÃtricos heteroscedÃsticos
• AperfeiÃoamento de testes de hipÃteses para modelos nÃo-lineares simÃtricos
• DiagnÃstico da gestÃo da manutenÃÃo na regiÃo metropolitana de Recife
• Modelo de apoio ao diagnÃstico no domÃnio mÃdico: aplicando raciocÃnio baseado em casos