The aim of this work was to evaluate arrays of fixed effects and residual covariance structures that best fit the herd and the animal variability to weights at birth, weaning, twelve and eighteen months of 3,690 Canchim animals. Different arrays of fixed effects (contemporary groups and, or the main effects of year, month or season of birth and sex) and different residual covariance structures (considering or not change of variance and of correlation between weights at different ages) were studied. The results indicated that the most adequate array of fixed effects to fit herd variability was the contemporary group of year, month and sex. The best residual covariance structures were the First-Order Factor Analytic and the Unstructured, which consider increase of residual variance as age increases and different correlation between pair of weights at different ages.
contemporary groups; longitudinal data; main effects; residual correlation
