Random regression models to estimate genetic parameters for fat and milk yield considering different residual variance structure

Abstract

Random regression models are an alternatively to adjust milk and fat yield test-day records along lactation curve. A total of 7,908 test-day records from 1,463 first lactation buffaloes were analyzed. The model included the additive genetic, permanent environmental and residual as random effects. As fixed effects the contemporary groups (herd, year-month of records), the linear and quadratic effect of age of cow at calving and the fixed curve of the population were considered. Residual variances were modeled trough a step function with 1, 4, 6 and 10 classes. Random effects were modeled through Legendre polynomials from third to sixth order. Residual variances were modeled with a step function with 4 classes. The models adjusting Legendre polynomials of fourth order for the additive genetic and permanent environmental (LEG4,4_4) and fourth and third order for the additive genetic and permanent environmental (LEG4,3_4), respectively, were the most adequate to described the trajectory of milk and fat yield, respectively. Milk yield heritability estimates obtained with LEG4,4_4 varied from 0.18 (first month) to 0.28 (10th month). Fat yield heritability estimates obtained with LEG4,3_4 varied from 0.20 (2nd month) to 0.27 (10th month). The residual variances for fat and milk yield should be modeled through heterogeneous classes, being four classes of residual variances the most adequate.