The objectives of this study were to evaluate a multibreed heteroskedastic animal model (MAMR) to estimate residual variance heterogeneity due to breed, heterozigosity, sex and contemporary groups (CG) and robustness to outliers. Differences on variance components, heritability and genetic values estimates were evaluated. It was used 4,016 records of post weaning weight gain (PWG) from Brangus foundation herd of Embrapa Pecuária Sul. Data were analyzed by MAMR with four different residual distribution specifications, which were defined by the nature of residual variance, homoskedastic (HO) or Heteroskedastic (HE) and by the Gaussian (G) or Student t marginal residual distribution. The best data fit was achived by using MAMR-T-HE and it was observed heteroscedasticity due to sex and animal heterozigosity and among contemporary groups. There were substantial differences in genetic variances of breed groups estimated by using different models and in the heteroskedastic, MAMR-G-HE and MAMR-T-HE. The greatest differences on heritability estimates occured among different breed groups. It was also noted a reordering in the ranking of superior animals regarded to genetic merit when heterogenous residual variances and robustness to outliers were considered. Multibreed genetic evaluations by using the Student t distribution model to describe residuals and assuming heterogenous residual variances are more appropriate to perform inference on genetic merit of animals from crossbreeding progenies of Angus and Nelore purebred animals.
Angus; Bayesian inference; beef cattle; crossbreeding; genetic evaluation; Nelore
