Statistical models to study of reproductive biotechnology in cattle

• Ano de defesa: 2023
• Autor(a) principal: Koeller, Andreza Jardelino
• 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: Binary data; Classificação de prenhez; Combined models; Dados binários; Embrião viável; Generalized linear mixed models; Machine learning; Modelos combinados; Modelos lineares generalizados mistos; Overdispersion; Pregnancy classification; Random forest; Superdispersão; Viable embryo
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11134/tde-12042023-170119/

The advent of artificial insemination and in vitro fertilization have made it possible for the field of animal breeding to gain sizeable advances in pregnancy outcomes. The data used in this work relates to embryo transfer and viability, where the nature of the variable under study are, respectively, binary and proportion samples. In this context, the goal of this work is to develop models capable of accommodating these kinds of data, and with data evaluate the possible influences for each of the interests, advancing knowledge in the field of statistics, as well as animal breeding. For the development of this work generalized linear mixed-effects models to evaluate the overdispersed binary data, with the objective of identifying which factors influenced a successful embryo transfer. Another goal was to verify which conditions lead to a high embryo viability rate, for that, combined models were proposed as a solution capable of accommodating the proportion data. Finally, in the last chapter, we proposed a comparison between different methodologies, which used the binary data with the objective of verifying the performance between statistical models with those proposed for machine learning.