Métodos de inferência para modelos de regressão aplicados a dados sorológicos de pacientes HIV+ com múltiplos níveis de censura à esquerda

• Ano de defesa: 2025
• Autor(a) principal: Felix, Matheus Henrique
• Orientador(a): Tomazella, Vera Lucia Damasceno
• Tipo de documento: Dissertação
• 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: Censura à esquerda; Modelagem estatística; HIV; LMD; Níveis de censura
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::REGRESSAO E CORRELACAO
• Link de acesso: https://hdl.handle.net/20.500.14289/21757

In this study, the focus is on data that exhibit the occurrence of left-censoring. This type of censoring occurs when the exact time of the event is unknown, but it is known to have occurred before the recorded time. A common example is in the analysis of serological laboratory data, which directly impacts decision-making by physicians, researchers, and other related specialists. Most published studies explore cross-sectional designs, where only one level of censoring is considered. However, occurrences of multiple left-censored measurements at different levels for a single patient are quite common when monitoring a patient's health over time. The same applies to tests conducted in different laboratories, as equipment varies, resulting in different censoring levels. Currently, researchers in the clinical field often exclude such data from their analyses. From a regression perspective, it is essential to adequately model the influence of time, factors, and/or covariates on viral load, as well as the correlation between repeated measurements for the same patient. These aspects are crucial for ensuring reliable statistical inferences, both in clinical trials and in observational cohort or case-control studies. In this context, the objective of this work is to present an appropriate methodology for handling data with multiple levels of left-censoring, particularly assuming Weibull and Log-Normal distributions, applied to real viral load data from HIV-infected patients.