Ajustes de modelos não lineares de efeitos fixos, com ponderação e misto

• Ano de defesa: 2014
• Autor(a) principal: Pereira, Glauber Márcio Silveira [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Biometria - Estudos longitudinais; Modelos não lineares (Estatistica); Estatistica matematica; Logística
• Link de acesso: http://hdl.handle.net/11449/113940

Nonlinear models tend to be used when it is suggested by theoretical considerations or by non-linear behavior. Even when a linear approximation is acceptable, a non-linear can also be used to obtain a clear understanding of the parameters. Nonlinear models have been applied to a variety of situations, even at finite populations. The objective of this research was: To study the Logistic model with adjustments in three errors structures: the fixed effects structure, structure with weighting and mixed-effects structure. It was used data of sweet orange fruit Citrus sinensis (L.) Osbeck in an experiment with five hearts grafted onto rootstock located at the Experimental Farm Lageado in Botucatu. Orange production was controlled by mass of fruits produced by plants in each crop annually in the period from April to May in the years 1969 to 1974, with the given mass of each plant in kg. To study of logistic, Gompertz, von Bertalanffy and Richards models with adjustments in three errors structures: the fixed effects structure, the weighting structure and mixed effects structure adjusted to data of volume of the trunk of Eucalyptus grandis, from three growing regions belonging to Votorantim Celulose e Papel. This region was chosen because favorably constitute the occurrence of rust of Eucalyptus and also because they are great for massive reforestation to restore degraded areas and for production purposes in both private companies. The criteria used for the selection of models were: Error mean square, Akaike information criterion, Akaike weight MOTULSKY & CHISTOPOULOS (2004), Bayesian Information Criterion Schwarz, Breusch Pagan test, Durbin Watson Test and squared correlation coefficient. To study the weight of fruit of the orange, the structure with weighting was more appropriate, since fixed the problem of heteroscedasticity and autocorrelation. For the data of volume of the trunk of Eucalyptus grandis, the structure with weighting was also ...