Identidade de modelos não lineares e regressão aleatória para o estudo da curva de crescimento de codornas de corte em diferentes gerações sob seleção

• Ano de defesa: 2014
• Autor(a) principal: Ribeiro, Jeferson Corrêa
• Orientador(a): Torres, Robledo de Almeida
• Banca de defesa: Nascimento, Moysés , Souza, Gustavo Henrique de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Doutorado em Zootecnia
• Departamento: Genética e Melhoramento de Animais Domésticos; Nutrição e Alimentação Animal; Pastagens e Forragicul
• País: BR
• Palavras-chave em Português: Codornas; Coturnix coturnix; Codornas - Aspectos genéticos; Codornas - Peso corporal; Polinômios de Legendre
• Palavras-chave em Inglês: quail; Coturnix coturnix; Quails - Genetic aspects; Quails - Body weight; Legendre polynomials
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::ZOOTECNIA
• Link de acesso: http://locus.ufv.br/handle/123456789/1887

The aim of this study was to compare the change of the growth curve by nonlinear regression models, using the test of identity of models, and by random regression models, using Legendre polynomials in a dataset with 15 generations of meat type quail in selection for weight increasing at 28 days.The dataset was divided in three groups of generations, initial (G1), intermediate (G2) and final (G3), each one with five generations, where the founder population contained 227 animals of UFV1 strain and 185 animals of UFV2 strain, resulting in a relationship matrix with 18,766 and 18,456 animals of both sexes and strains.For the study through nonlinear regression models were used five body weight records, where animals besides birth weight, were individually weighed at 7, 14, 21 and 28 days of age.Initially, it was sought to identify the best nonlinear model describing the growth curve models using the von Bertalanffy, Brody, Gompertz, Logistic, Richards and segmented linear regression (spline) with a knot at 14 days.For the study of the growth curve through random regression models, individual weights were analyzed using the animal model, where ages were described in terms of orthogonal Legendre polynomials, with generation and sex as fixed effects.Among the analyzed models, the Gompertz model was the most appropriate to describe the growth curve of quails for both sexes in all groups of generations. From the best model that was identified, it was possible to analyze identity models to see if there were changes in the growth curve of the animals.The result of the likelihood ratio tests for model identity indicated that the adult weight and maturing rate were significantly different between the three groups of generation, for both sexes, not being possible to use a single model that describes the growth in the three groups of generations.In random regression models, phenotypic variance estimates were similar for the performance in three groups, in both strains. However, the additive genetic variance decreased over the generations, while there was an increase of the variance of permanent environmental and residual, as the selection acted.The heritability (h2) decreased over the generations, being high at the birth weights, reducing the intermediate ages and increasing as the animal approached adulthood.Thus, it is concluded that there was a change in the growth curve of the animals over the generations, with an increase in asymptotic weight and reduction of animal maturity, demonstrating that the selection process was efficient.Selection of animals at 28 days of life is recommended, because this is the age with the highest value of estimated h2. Because there are differences between groups of generations, it is also recommended to use only data from the last generations for traits related to the animal weights instead of using all the information contained in the data set, especially when there are large numbers of animals in initial generations, for both nonlinear regression models, as for random regression, thereby reducing estimation errors.