The objective of the present study was to analyze models with genetic and/or residual heteroskedasticity for genetic evaluation of the weaning weight of Charolais-Zebu crossbred calves. Weaning weight data from 56,965 crossbred calves were analyzed using animal models with different combinations of genetic and residual heteroskedasticity and/or homoskedasticity. The inference on a posteriori distributions of genetic parameters were by the Monte Carlo method via Markov chains. The model with genetic and residual heteroskedasticity was the best fit on the data. Groups of animals with different genetic compositions, expressed as percentages of Charolais-Zebu breed alleles and individual and maternal heterozygosis, had different genetic variances. These genetic variances could be modeled by linear functions of the Charolais and Zebu genetic variances and the variance attributed to segregation. The breed compositions, the individual and maternal heterozygosis, the sex and age of dam at calving were significant sources of residual heteroskedasticity. The a posteriori means for heritabilities and sire and dam classifications were altered due to genetic and residual heteroskedasticity.
animal breeding; Bayesian inference; Canchim; crossbreeding; selection
