Comparação de ajustes do modelo de Gompertz a dados de crescimento

• Ano de defesa: 2013
• Autor(a) principal: Sallum Neto, Farid [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Biometria; Modelos matematicos; SAS (Sistema de computador); Modelos não lineares (Estatistica); Mathematical models
• Link de acesso: http://hdl.handle.net/11449/108592

In growth model analyzing to represent the development of organisms for a larger period than its initial phase, it nonlinear models are often used . The aim of this work was to adjust the nonlinear Gompertz model using the software SAS by three different structures: 1) fixed effects model, 2) first-order autoregressive model and 3) mixed-effects model to a data of growth of female and male rats Rattus norvegicus and to a data of growth of three breeds cow to verify the best model for the data obtained. To attach the objective, six criteria was chosen as evaluators of statistical adjustment. The criteria were: residual mean square, Akaike information criterion (AIC), Bayesian information criterion (BIC), mean prediction error (EPM), residues in starting points, called res 0, used to assess the fit in the first observations and an index value being considered as an evaluator at the end of adjustment (calculated as the ratio between 20% of the largest estimated values and estimated asymptote). In addition, Durbin-Watson and Breusch-Pagan tests were performed to verify the independence of waste and heteroscedasticity, respectively. The results show that in many repetitions, where was need a correction of autocorrelation and heteroscedasticity, the first order autoregressive model was the best fit. In relation to the single adjustment, where the autocorrelation was significant, the autoregressive model was the best, according to the criteria utilized, and when the autocorrelation was not significant, the fixed effects model was the best model