Random regression models using Legendre polynomials were used to describe the growth curve of Nelore cattle weighted every three months from birth to 720 days of age, in Pernambuco state, northeastern Brazil. Six different random regression models using Legendre polynomials of three, four, and five degrees to model additive genetic and permanent environmental effects, under homogeneous and heterogeneous residual variances with three classes were evaluated. According to the Akaike's information criteria, the five degree Legendre polynomial with homogeneous error was the best fitting model. Genetic and phenotypic correlations were low between weights at early and late stages of the weighting period, and high and approximately constant between weights at late stages. Genetic covariance of weights increased with age. Heritability estimates were low to moderate up to 60 days of age and reached values between 0.50 and 0.60 at older ages.
Akaike's information criteria; Legendre polinomial; likelihood ratio test; random regression
