Aplicações das distribuições weibull modificada e Beta-Weibull na presença de frações de cura sob o enfoque Frequentista e Bayesiano

• Ano de defesa: 2016
• Autor(a) principal: Peres, Marcos Vinicius de Oliveira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Programa de Pós-Graduação em Bioestatística; UEM; Maringá, PR; Centro de Ciências Exatas
• Palavras-chave em Português: Métodos quantitativos aplicados à saúde, Weibull modificada , Frequentista e Baysiano, Comparação, Métodos quantitativos aplicados à saúde, Beta-Weibull, Frequentista e Bayesiano, Comparação, Análise de sobrevivencia (Bioestatística), Função de risco, Intervalo de confiança perfilado - Brasil; Survival analysis, cancer, profile likelihood confidence interval, hazard function - Brazil; Ciências Exatas e da Terra; Estatística
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/4357

The medical research is of great importance for the monitoring of patients who have cancer - especially after medical interventions, such as surgical resections, organ transplants and chemoradiotherapy - aiming a better understanding in the treatment and a life quality improvement for the subjects. It is extremely important to use proper methods for the data modelling. Among the available tools, a very important one is the survival analysis. In the survival analysis context, the event of interest is often related to death or disease recurrence. However, in the study´s conclusion it is possible for a part of the sample not to suffer from the event of interest. These patients can have been cured or be immune to the event of interest. Therefore, it is quite important to estimate in a proper way the proportion of not susceptible patients. The traditional models, which are well known in the survival analysis, usually are not adequate to estimate the immune proportion. It is necessary the use of complex statistic models to incorporate this information. Currently there are several methods to estimate the immune proportion, such as the mixing models with cure fraction and the not mixing models. In the present paper it has been used an analysis based on two not common distributions in the practical context, the Weibull modified distribution, a distribution composed by three parameters. Likewise, the beta-Weibull distribution, with four parameters. For both distributions, it was considered the cure fraction existence, censored data and the covariant. Frequentist estimates (by maximum likelihood) and Bayesian inference estimates were compared. To verify the adequacy of these models in the real data analysis the Weibull modified analysis was applied to the data of patients who have gastric adenocarcinoma. And the beta-Weibull analysis for the data of bone marrow transplant. Both considered models have adjusted in a satisfactory way to the data and properly estimated the cure proportion. The Bayesian estimates and their respective High Posterior Density intervals (HPD) were more parsimonious than the ones resulted from the method of maximum likelihood