Models based on skewed distributions for paired data, with applications to health data

• Autor(a) principal: Cunha, Guilherme Zubatch da
• Data de Publicação: 2020
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/17/17139/tde-23082020-143052/
• topic: Análise de Variância; Analysis of variance; Bioestatística; Biostatistics; Dados em saúde; Dados longitudinais; Dados pareados; Health data; Longitudinal data; Modelos de regressão; Paired data; Regression models

This thesis aims to introduce and raise interest on pretest/posttest analysis in epidemiological studies. In a first article, conceptual aspects are described and methodological guidelines for this type of analysis are systematized. We defined the key concepts of pretest and posttest and how they can be applied to epidemiology. Methods to assess pretest and posttest are very common and there are many ways to resolve the same problem. The objective here is to present some of them: t-test, analysis of variance (ANOVA) on the gain scores, analysis of covariance (ANCOVA), ANOVA on residual scores, and repeated measures ANOVA. The second article presents a general family of skew-symmetric distributions applied on pretest/posttest in epidemiological studies by considering a longitudinal approach. Situation where the normality assumption is not valid are very common, especially when data present a skewed distribution. For taking into account possible skewness effects, we propose the use of longitudinal analysis from a Bayesian perspective. In both articles, results of an application to a real data set related to multiple sclerosis are reported.