• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.11.2022.tde-06042022-104534
Document
Author
Full name
Allison Queiroz de Oliveira
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
Piracicaba, 2022
Supervisor
Committee
Lima, Cesar Goncalves de (President)
Giannotti, Juliana Di Giorgio
Medeiros, Simone Daniela Sartorio de
Title in Portuguese
Modelagem de peso e consumo de sólidos totais em bezerros leiteiros e o uso dos resíduos de confundimento mínimo no diagnóstico de modelos lineares mistos
Keywords in Portuguese
Correlação
Dados longitudinais
Desempenho
Modelos hierárquicos
Abstract in Portuguese
O uso de modelos mistos na análise de respostas influenciadas por fatores de tratamento e por fatores longitudinais, apresenta muitas vantagens quando comparado aos modelos clássicos de regressão. A complexidade desses modelos aumenta quando se modela simultaneamente duas ou mais variáveis respostas correlacionadas, num cenário com forte desbalanceamento no número de observações. No diagnóstico dos modelos mistos ajustados sugere-se o uso dos resíduos de confundimento mínimo em substituição aos resíduos clássicos dos modelos de regressão. No presente trabalho foram utilizados os dados oriundos de um experimento desenvolvido na Escola Superior de Agricultura “Luiz de Queiroz”, que avaliou o desempenho de bezerros da raça holandesa, submetidos a três dietas sólidas ao longo de oito semanas. Objetivou-se o ajuste de modelos lineares mistos sob as abordagens univariada e bivariada para explicar o comportamento das variáveis respostas peso e consumo médio de sólidos totais ao longo do tempo, utilizando os resíduos de confundimento mínimo no diagnóstico dos modelos selecionados. Nos ajustes foram utilizadas diversas bibliotecas do programa computacional R e na seleção dos modelos foram utilizados testes da razão de verossimilhanças para modelos encaixados e o Critério de Informação Bayesiano (BIC) para os modelos não encaixados. A abordagem bivariada foi mais adequada e informativa do que a univariada, porque considera a correlação entre o peso e o consumo medidos nas mesmas unidades experimentais e nos diferentes instantes do tempo. Comprovou-se a eficácia do uso dos resíduos de confundimento mínimo no diagnóstico dos modelos lineares mistos em comparação aos resíduos marginais e condicionais estudentizados, visto que os primeiros apresentam um menor viés quando comparado aos dois últimos.
Title in English
Modeling weight and total solids consumption in dairy calves and the use of least confounded residuals in the diagnosis of linear mixed models
Keywords in English
Correlation
Hierarchical models
Longitudinal data
Performance
Abstract in English
The use of mixed models in the analysis of responses influenced by treatment factors and by longitudinal factors presents many advantages when compared to classical regression models. The complexity of these models increases when two or more correlated response variables are modeled simultaneously, in a scenario with a strong imbalance in the number of observations. In the diagnosis of adjusted mixed models, the use of least confounded residuals is suggested to replace the classical residuals of regression models. In the present work, data from an experiment developed at the “Luiz de Queiroz” College of Agriculture were used, which evaluated the performance of Holstein calves submitted to three solid diets over eight weeks. The objective was to adjust linear mixed models under univariate and bivariate approaches to explain the behavior of the response variables weight and average consumption of total solids over time and the use of least confounded residuals in the diagnosis of the selected models. In the adjustments, several libraries of the computer program R were used and in the selection of the models, tests of the likelihood ratio were used for nested models and the Bayesian Information Criterion (BIC) for the not nested models. The bivariate approach was more adequate and informative than the univariate one, because it considers the correlation between weight and consumption measured in the same experimental units and at different moments in time. The effectiveness of the use of least confounded residuals in the diagnosis of linear mixed models was proved in comparison to the marginal and conditional studentized residuals, since the first one have a lower bias when compared to the last two.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2022-04-07
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.