Considerations about nonlinear regression

Differently from linear regression models, where the quality and validity of a proposed model are checked through regression diagnostic methods, for nonlinear models, we should consider additional diagnostic methods. These special diagnostic methods for nonlinear regression models measure the nonlin...

ver descrição completa

Detalhes bibliográficos
Autores: Mazucheli, Josmar, Achcar, Jorge Alberto
Formato: artículo
Estado:Versión publicada
Fecha de publicación:2008
País:Brasil
Recursos:Universidade Estadual de Maringá (UEM)
Repositorio:Acta scientiarum. Technology (Online)
Idioma:portugués
OAI Identifier:oai:periodicos.uem.br/ojs:article/2551
Acesso em linha:http://www.periodicos.uem.br/ojs/index.php/ActaSciTechnol/article/view/2551
Access Level:acceso abierto
Palavra-chave:modelos de regressão não linear
medidas de curvatura
medida de vício
métodos de simulação
medidas de assimetria
1.02.02.00-5 Estatística
Descrição
Resumo:Differently from linear regression models, where the quality and validity of a proposed model are checked through regression diagnostic methods, for nonlinear models, we should consider additional diagnostic methods. These special diagnostic methods for nonlinear regression models measure the nonlinearity amount of the proposed model. Nonlinear models with behavior too different from linear models could invalidate the asymptotical results, used to get interest inferences, especially in small-sized samples. In this paper, an overview of the main nonlinearity checking methods is offered. In particular, Bates and Watts curvature measures, Box bias measure, simulation methods use and some nonsymmetry measures are focused.