Modelo de regressao Weibull discreto com fracao de cura em dados de sobrevivencia

Abstract

This paper presents a regression model for discrete time data with cure fraction. For this purpose, we considered a mixture model, in which times are modeled through the discrete Weibull distribution and the cure probability modeled with covariates by using the logit link function. Considering that the model is a mixture, the estimation of the parameters was performed by EM algorithm. The behavior of the estimating algorithm was tested with Monte Carlo simulation experiments and an application for real data was added.

Keywords

• Modelo de mistura
• regressao
• distribuicao Weibull discreta
• algoritmo EM