Data provided by the Brazilian Zebu Breeders Association (ABCZ), consisting of 21,762 records from 4,221 animals of Tabapuã cattle, weighted from birth to 550 days of age, were used to estimate covariance functions by random regression models using Legendre polynomials of order two to five. Models included the direct and maternal genetic, animal and maternal permanent environmental random effects and compared by Schwarz´s Bayesian information criteria (BIC) and Akaike´s information criteria (AIC). Both criterions suggested the model including direct genetic, maternal genetic, animal permanent and maternal permanent environmental effects respectively adjusted by cubic, quadratic, fourth order and linear polynomials, and residual variances adjusted by fifth order variance function as the best one to describe the covariance structure of the used database. Direct heritability estimates were higher at the beginning and at the end of the growth trajectory. Maternal heritability estimates increased from birth to 160 days of age and decreased thereafter. In general, genetic correlation estimates decreased as age between weights increased. Efficiency of selection may be improved by using weights of the post weaning period because of their higher genetic variance and heritability estimates.
beef cattle; covariance function; growth curve; longitudinal data
