Estimation of genetic parameters for growth evaluation of the Nelore cattle created in Brazilian north region using random regression models

This study aimed to compare different random regression models and determine the most appropriate to describe changes in growth assessment parameters of Nellore cattle raised in northern Brazil. Were evaluated 32,543 weight records from birth to 660 days of age. For adjustment of the average regression curve were used Legendre polynomial with orders ranging from two to eight for fixed effects modeling. The best fit was cubic degree, according to the criteria: mean square error (MSE), the coefficient of determination (R?), mean absolute deviation (DMA) and Percentage of Squared Bias (PSB). For random regression models, it was considered as fixed effects the contemporary groups and age at calving, linear and quadratic effects. Random effects were additive genetic and permanent environment, with both animal and maternal effects; and residual effect. These models were compared using the log-likelihood function, the Akaike information criteria and Bayesian Schwarz. A model using fifth-degree polynomials for additive genetic animal and maternal effects, and two-degree polynomials for permanent environmental animal and maternal effects and 7 classes of residual variance was the best fit one. Estimates of direct additive genetic variance increased by age, and the maternal additive genetic variance showed varied pattern. Estimates of direct heritability showed a slight decrease at first (0.15) and increased from 40 days to the following ages. Approximately at 100 days the heritability gradually increases from 0.35 to 0.6 at 660 days. Maternal heritability presented a growth peak from birth to 90 days of age, decreased abruptly until near to 370 days, when it was maintained low values at later ages. The response to selection to obtain heavier animals will be effective when performed at from 100 days on when the values of direct heritability were of medium to high magnitude.