Modelos lineares generalizados mistos e equações de estimação generalizadas para dados binário aplicados em anestesiologia veterinária

• Ano de defesa: 2015
• Autor(a) principal: Galdino, Maicon Vinícius [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Anestesia veterinaria; Anestesiologia; Equino - Doenças - Tratamento; Analgesicos; Modelos lineares (Estatistica); Distribuição binomial
• Link de acesso: http://hdl.handle.net/11449/132051; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/02-10-2015/000850580.pdf

The objective of this work is to propose, compare and select statistical models based on the methodology of GLMMs and GEEs. Examine in which situations each model can be better used. Discuss about the techniques of parameter estimation, choice of the covariance matrix structure, definition and utility of the odds ratios and application in a set of data in the area of veterinary anesthesiology involving equines. The data used in this work were obtained by Taffarel (2013). The twentyfour equines used in the experiment were grouped as follows: (1) anesthetized animals (treatment 1), (2) animals anesthetized and that received prior analgesia (treatment 2), (3) animals anesthetized and underwent orchiectomy with administration of analgesics after surgery process (treatment 3) and (4) animals anesthetized and underwent orchiectomy with administration of analgesics before being operated (treatment 4). The animals were observed by five evaluators at different times: one hour before surgery or anesthetic (M1), four hours after recovery from anesthesia and before the application of analgesics in animals of treatment 3 (M2), two hours after M2 (M3) and twenty-four hours after surgery (M4). The response variable (with Bernoulli distribution) was look the flank, a common behavior in equines with abdominal pain. The treatment 3 and the moment 2 were the effects with the highest occurrence of odds ratios of the response variable (regardless of what statistical methodology used in the analysis) compared to other treatments and moments. Both statistical models fitted to the data (models 1 and 2) obtained through the EEGs and MLGMs shown to be a powerful tool to explain the response variable look flank especially regarding the selection of statistically significant covariates. In the EEG, is not required as the specification of the distribution of vectors response, it became quite advantageous statistical methods, particularly for binary data as ...