The parametric and semiparametric regression models based on the generalized odd log-logistic family

• Ano de defesa: 2020
• Autor(a) principal: Prataviera, Fábio
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Censored data; Cubic spline; Cure rate; Dados censurados; Fração de cura; Inflação de zeros; Simulação; Simulation; Spline cúbico; Zero inflated
• Link de acesso: http://www.teses.usp.br/teses/disponiveis/11/11134/tde-04062020-122642/

In this work, several analyzes were performed through regression models considering the family of new distributions, called generalized odd log-logistic-G (GOLL-G), the distributions in this family have greater flexibility, such as functions of bimodal densities. Based on the GOLL-G family, we proposed: regression models with different regression structures; inflated semi-parametric model of zeros modeling of the parameters via penalized splines; For all the modeling approaches presented, the computational resource for the implementation of the models was software R, throughout the document as well as brief descriptions of the codes used. The results obtained in the applications show that the proposed model can be an interesting alternative, especially when the data present asymmetry and bimodality.