Comparison of models in genetic evaluation of growth traits of Brahman cattle in southeastern Mexico

Variance components and genetic parameters were estimated for birthweight (n=2905), weaning weight adjusted to 205 d (n=2264), and preweaning daily gain (n=2264) in four herds of Brahman cattle in southeastern Mexico. Six animal models were fitted for each trait, ranging from a simple model, with the additive direct effect (σ2a) as the only random effect, to the complete model with σ2a, additive maternal effect (σ2m), covariance between them (σam) and maternal permanent environmental effect (σ2c). The six models included as fixed effects contemporary group of birth (herd-year-season), sex and parity number. The likelihood ratio test proved that for all traits σ2m and σ2am were important (P <0.05), but σ2c was not (P >0.05). Ignoring maternal effects in the models tended to increase direct heritability, while ignoring σam tended to increase error. Elimination of σ2c did not affect direct and maternal heritability estimates. Spearman correlations of direct and of maternal animal genetic values, between models with and without σ2c were 1, suggesting no differences in the ranking of animals. The genetic correlations between additive and maternal heritabilities for all characteristics were negative, which affected estimated total heritability (lower than in models omitting σam). The additive genetic variance, given maternal effects (σ2a.m), was constant across models suggesting a correct partition of additive genetic effects into their direct and maternal components. In conclusion, the most appropriate model must be selected for each characteristic, in order to reduce bias in genetic evaluation.