Modelos de sobrevivência induzidos por fragilidade discreta com fração de cura e riscos proporcionais

• Ano de defesa: 2022
• Autor(a) principal: Espírito Santo, Ana Paula Jorge do
• Orientador(a): Cancho, Vicente Garibay
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa Interinstitucional de Pós-Graduação em Estatística - PIPGEs
• Palavras-chave em Português: Fragilidade discreta; Fração de cura; Riscos proporcionais; Distribuição Katz; Distribuição Poisson Generalizada
• Palavras-chave em Inglês: Discrete Frailty; Cure rate; Proportional hazards; Katz distribution; Generalized Poisson distribution
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::INFERENCIA PARAMETRICA
• Link de acesso: https://repositorio.ufscar.br/handle/ufscar/17156

This work presents two new survival models induced by discrete frailty with unobserved heterogeneity and proportional hazards structure, for lifetime data. The first model consider the discrete frailty variable with Katz distribution and the second with Generalized Poisson distribution, which have overdispersion, equidispersion and underdispersion properties. The new models encompasses as particular case the promotion cure rate model. The proposed model with katz discrete frailty also encompasses the mixture cure rate model and cure rate model with dispersion. Inference aspects for proposed models as discussed, in a classical approach for Katz distribution, for which the maximum likelihood tools were used and regression models were presented to evaluate the effects of covariates in the cured fraction. Furthermore, an expectation maximization algorithm for determining the maximum likelihood estimates of the parameters of the model was presented. For Generalized Poisson model, a baysean aprouch was also used, through Markov chain Monte Carlo simulation method, especifically Metropolis–Hastings algorithm. Finally, the modeling was fully illustrated on cervical cancer data sets.