Uso de polinômios segmentados na modelagem de dados longitudinais de ponderal em bovinos da raça Tabapuã

• Autor(a) principal: Menezes, Gilberto Romeiro de Oliveira
• Data de Publicação: 2010
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/1314
• topic: Curva de crescimento; Regressão aleatória; Parâmetros genéticos; Bayesiana; Growth curve; Random regression model; Genetic parameters; Bayesian; CNPQ::CIENCIAS BIOLOGICAS::GENETICA::GENETICA ANIMAL

Weight records from the 1st to 600th day of age of 84,215 animals of the Tabapuã’s breed were used in order to evaluate the use of linear B-splines functions in the modeling of longitudinal weight data using random regression model (RRM). To compare the results obtained with RRM, multi-trait analysis was performed with standard pre-adjusted weights at 120, 240, 360 and 480 days. A RRM was applied to a data file similar to the one used to multi-trait analysis (same animals and weight records, but without pre-adjust for the evaluated ages - RRM1) and another was applied to complete data file (same animals, but with all available weight records from these animals - RRM2). RRM1 and RRM2 included direct and maternal additive genetic and direct and maternal permanent environment random effects, differentiating by the number of residual classes, four to RRM1 and six to RRM2. Six knots located at points 0, 120, 240, 360, 480 and 600 days were considered for RRM1 and RRM2. The estimation of (co) variance components for the three models was performed using Bayesian approach via Gibbs sampling. In general, the three models generated (co) variance components and genetic parameters similar for weight at 120, 240, 360 and 480 days. RRM1 presented problems in the modeling of the extremes of the range of data evaluated. Regarding the classification by breeding values, the models also generally were similar, with larger differences in the assessment of animals with less information. Considering the results, and, for allowing better use of available data and for not requiring pre-adjustment at data, the use of linear B-spline functions in RRM should be recommended for use in genetic evaluation of longitudinal weight data in beef cattle.