Statistical modelling and computational tools applied to the natural sciences

• Ano de defesa: 2020
• Autor(a) principal: Fatoretto, Maíra Blumer
• 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: Análise bivariada; Bivariate analysis; Bootstrap não paramétrico; Correlações; Correlations; Generalized linear mixed models; Modelos lineares generalizados mistos; Non-Parametric Bootstrap; Overdispersion; Superdispersão
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11134/tde-11012021-182812/

Natural sciences data can involve different characteristics, making it very interesting to be analyzed and may bringing many contributions from a statistical point of view. One of its ma- jor areas is entomology, a science that studies insects and where we can find proportions, count or continuous data, univariate or multivariate responses, in addition to repeated measures. In this sense, to develop models and computational methods that support these observations and to bring gains to the statistical area, at the same time as making this analyzes more accessible is the aim of this work. Generalized linear mixed models were used to analyze proportion and overdispersed data in dose-response studies and, the goodness-of-fit was assessed using half-normal plots with simula- tion envelopes. To compare treatments of these data different non-parametric bootstrap resampling methods were proposed and, a simulation study was carried out to verify the performance of different types of confidence intervals. Besides, an extension of the bivariate normal model was developed to accommodate count data. This extension was developed by adding a new parameter in the variance- covariance matrix to better accommodate over- or underdispersion for count data modeling the mean and dispersion of the observed data.