Análise de agrupamento para a avaliação de identidade de modelos não-lineares em análise de sobrevivência

• Autor(a) principal: Tomaz, Flávia Sílvia Corrêa
• Data de Publicação: 2009
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/4017
• topic: Ward; Análise de sobreviência; identidade de modelos; Ward; Survival analysis; Identity of models; CNPQ::CIENCIAS AGRARIAS

The aim of this study was to compare non-linear models fitted to data on survival of ants under different treatments using alternative methodology. This methodology consists of using the technique of cluster analysis, Ward's method, to evaluate the identity of models used in survival analysis. The data used in this work are related to an experiment conducted in the entomology laboratory of the Federal University of Viçosa, Brazil. It also used a set of simulated data based on the Weibull distribution. Originally was applied a nonparametric technique, Kaplan- Meyer estimator, to estimate the survival curves, then the logrank test for comparison of these curves. For the real dataset it was fitted a logistic model of survival times, while for the simulated data it was fitted a Weibull model. The estimates of the parameters of each fitted model were grouped using the technique of cluster analysis. The results found by the grouping were equivalent to that by the logrank test. It is concluded that the proposed methodology showed to be efficient and less laborious, mainly when many survival curves need to be compared.