Parâmetros genéticos para produção de leite no dia do controle de vacas da raça holandesa criadas no Rio Grande do Sul

• Autor(a) principal: Dornelles, Mariana de Almeida
• Data de Publicação: 2014
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• dARK ID: ark:/26339/001300000ghp1
• Texto Completo: http://repositorio.ufsm.br/handle/1/28462
• topic: Componentes principais; Funções paramétricas; Polinômios de legendre; Posto reduzido; Principal components; Parametric functions; Legendre polynomials; Reduced rank; CNPQ::CIENCIAS AGRARIAS::ZOOTECNIA

This study aimed to estimate genetic parameters for milk production on the control day of primiparous Holstein cattle raised in Rio Grande do Sul, through random regression models. In Article 01, the test day milk production (PLDC) was grouped into ten monthly classes of lactation, obtained between the 5th and the 305th days postpartum (PLDC1 to PLDC10). Analyses considering 11 different models were conducted: multi-characteristic pattern (MC), five reduced rank models adjusting the first main components (m = 1, 2, 3, 4 and 5) for the direct additive genetic effect and five models using analysis of factors (m = 1, 2, 3, 4 and 5). For the PLDC, the linear model included the effects of age at calving (linear and quadratic) and the number of days in lactation as covariates in addition to the contemporary group as a fixed effect. According to the comparison criteria, the model that adjusted the first four principal components (CP4) is the one that has provided the best fit. Estimates of Phenotypic co (variances), direct additive genetic, of permanent and residual environment obtained using the MC and CP4 models were similar. Direct heritability estimated for the ten PLDC using the MC, CP4 and AF4 models were similar and ranged from 0.06 (PL6) to 0.65 (PL10). Estimates of genetic and phenotypic correlations obtained by MC and the CP4 were equal. The reduced rank model reduced the number of parameters in the analysis, without reducing the quality of the fit. In Article 02, the PLDC of primiparous cattle were grouped into biweekly classes of lactation, ranging from 1 to 20 classes, Class 1 consisting of lactation measures between day 6 and 20, and class 2, between day 21 and 35, successively. Initially, the residual variance was considered homogeneous throughout lactation, subsequently they were assumed heterogeneous between the groups and homogeneous within each group. When considering homogeneous residual variance, it was found, according to -2LogL, AIC and BIC, that the model that used the function of Ali & Schaeffer (FAS) provided a superior fit to model the trajectory of additive genetic and permanent environmental variances of the PLDC, compared to the one that used the Wilmink function (FW). However, in this study, the superiority of parametric functions with respect to Legendre polynomials was observed only when the FW was used, ie, by using Legendre polynomials of the same order as the FAS, it was possible to observe better values of AIC, BIC and -2LogL for the Legendre polynomial model of order 5 (LEG5_HO). The Legendre polynomial of quintic order was more appropriate than the function of Ali & Schaeffer for genetic studies of milk production in the control day of Holstein cattle. The model that best fit the production of milk in the control day was the one that considered 20 classes of heterogeneous variance. However, as there are classes with similar residual variances, it is possible to group them and reduce the number of estimated parameters, decreasing the computational requirements for the adjustment of the models. In Article 03, the PLDC of primiparous cattle were grouped in to fortnightly classes of lactation, ranging from 1 to 20 classes in which Class 1consists of measures of lactations between 6 and 20, class 2, between 21 to 3, subsequently. Initially, analyses were performed considering 13 different models, different orders of adjustment of orthogonal Legendre polynomials, both for the direct genetic effect (m =3, 4, 5 and 6) and for the permanent environmental effect (m =3, 4,5, 6 and7). It was performed 13 analyses considering different types of reduced rank, setting the first principal component (m = 1, 2, 3, 4 and 5) to direct additive genetic effect. The results indicated that only four principal components are required to model the structure of (co) variance among dairy genetic control, reducing the number of parameters in the analysis. When comparing the model of full rank (LEG_67) to the model of reduced rank (CP46 ), it was observed a similar behavior in all estimates of variances. The estimated heritability for the two models were very similar for all PLDC and showed, as expected, the same trend of the variance components of random genetic additive effects, with higher values at the extremes of the curve. Estimates genetic correlations values in the models refer to the correlation of the tenth week of lactation with the others, and ranged from 0.32 to 0.99 in the first half to the ninth half of lactation, in other words, the measure decreased as PLCD moved away in time.