Componentes de covariâncias estimados por metodologia bayesiana para parâmetros biológicos obtidos por modelos não lineares para bubalinos da raça Murrah

Abstract

The aimed of this work was to study the adjustment of classical non linear models, Von Bertalanffy, Brody, Gompertz, Logistic to growing records of buffaloes of Murrah breed, raised on lowlands in the State of Rio Grande do Sul, and to estimate covariance components by Bayesian focus, for growing curve parameters with biological interpretation. In paper 01 there were studied the adjustment of the classical non linear models already mentioned to growing data for a group of 66 buffaloes females, born from 1982 to 1989, sired by three males and 38 females. There were evaluated the traits Asymptotic weight (A) and Maturity rate (K). The total pair of records weight-age was 26 weighting/female and 1,638 observations. The criterions utilized to select the model that better adjust the growing curve were: asymptotic standard deviation (DPA); the determination coefficient (R2); the residual absolute average deviation (DMA) and asymptotic index (AI). It was concluded that all the models overestimated the birth weight (PN) in bigger or smaller magnitude. In crescent order, the models Von Bertalanffy, Gompertz, Logistic and Brody overestimated PN by 28.55; 32.74; 42.70 and 43.45 kg, respectively. The Logistic model underestimated A (-2.09 kg) and Von Bertalanffy, Gompertz, and Brody overestimated A in crescent order 8.04, 17.7 and 280.33 kg, respectively. Based on the adjustment criterions and in the predicted curves behavior, the Gompertz model, followed by Logistic and Von Bertalanffy were the best adjustment. In Paper 2 there were studied the adjustments of the same models and for the same traits in Paper 01 for a group of 67 buffaloes, born from 1982 to 1989 sired by three males and 42 females. It was concluded that all the models overestimated PN. Von Bertalanffy and Brody models overestimated A, and Gompertz and Logistic models underestimated it. The smaller DPA was obtained by Brody model characterizing a bigger R2 but this model presented the bigger DMA. Considering all the criterions, Gompertz model presented the best adjustment followed by Logistic and Von Bertalanffy. It is suggested do not use Brody model to describe the growing curve for animals of Murrah breed raised in the conditions of this work. In Paper 3 there were estimated covariance components and genetic parameters by Bayesian focus, using the Family BLUPF90, for the parameters A and K, estimated by Gompertz model and adopting an animal model. The heritability coefficients presented elevated values for A and for K (0.57 and 0.34, respectively), indicating that selection can be used as an instrument for change the curve shape of this population. However, the use of this information must be done with to much attention because these traits are negatively correlated. In this case a restricted selection index should be used with more success.