Modelo de mixtura Logit-Weibull para datos censurados con fracción de cura
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Editor
Pontificia Universidad Católica del Perú
Acceso al texto completo solo para la Comunidad PUCP
Resumen
Los modelos de mixtura con fracción de cura se propusieron originalmente en el área de
medicina, para modelar la curva de supervivencia a largo plazo de pacientes con cáncer que
llevaban diferentes tratamientos. En la presente tesis se ha presentado y desarrollado el mode-
lo de riesgos proporcionales con fracción de cura Logit Weibull (PHMC-Logit-Weibull) para
tiempos censurados por la derecha. De manera específica, este modelo asume que la población
está compuesta por dos sub poblaciones llamadas susceptibles y no susceptibles. Este modela
la probabilidad de ser o no susceptible por un modelo de regresión logística y el tiempo a
la ocurrencia del evento de interés para la población susceptible por un modelo de riesgos
proporcionales con función de riesgo basal Weibull. La estimación de los parámetros se da vía
inferencia clásica con métodos de optimización numérica y el uso del algoritmo de Esperanza
y Maximización (EM). En el estudio de simulación se pudo corroborar que el proceso de es-
timación se realiza de forma adecuada independientemente del tamaño de muestra y número
de covariables.
Mixture models with fraction of cure were originally proposed in the eld of medicine to model the long-term survival curve of cancer patients undergoing dierent treatments. In this thesis, the proportional hazard cure fraction model Logit-Weibull model has been presented and developed for right censored data. More specically, this model assumes that the popula- tion is composed of two sub-populations called susceptible and non-susceptible and therefore, the probability of being or not susceptible can be modeled by a logistic regression model and the time to the occurrence of the event of interest by a proportional hazards model with baseline Weibull risk function. The estimation of the parameters is given by classical infe- rence with numerical optimization methods using the Expectation and Maximization (EM) algorithm. In the simulation study, it was possible to corroborate that the estimation process is carried out adequately regardless of the sample size and number of covariates.
Mixture models with fraction of cure were originally proposed in the eld of medicine to model the long-term survival curve of cancer patients undergoing dierent treatments. In this thesis, the proportional hazard cure fraction model Logit-Weibull model has been presented and developed for right censored data. More specically, this model assumes that the popula- tion is composed of two sub-populations called susceptible and non-susceptible and therefore, the probability of being or not susceptible can be modeled by a logistic regression model and the time to the occurrence of the event of interest by a proportional hazards model with baseline Weibull risk function. The estimation of the parameters is given by classical infe- rence with numerical optimization methods using the Expectation and Maximization (EM) algorithm. In the simulation study, it was possible to corroborate that the estimation process is carried out adequately regardless of the sample size and number of covariates.
Descripción
Palabras clave
Modelos de riesgos proporcionales, Distribución de Weibull, Cálculo fraccional, Análisis de regresión logística