Comparison of models and impact of missing records on genetic evaluation of calving ease in a simulated beef cattle population

Abstract

This study compared the application of a bivariate linear-linear (LL) and a linear-threshold (LT) sire-maternal grandsire model for genetic evaluation of calving ease (CE), using birth weight (BW) as a correlated trait, and assessed the impact of missing records on genetic evaluation of CE in a simulated multi-breed beef population that mimicked phenotypic and genetic parameters of beef cattle in Ontario. Models included fixed age-of-dam by sex-of-calf, management group, breed and heterosis effects, and random direct and maternal genetic, maternal permanent environment and residual effects. The LL model was applied to BW and CE Snell scores, and LT model was applied to BW and CE raw scores. CE evaluations were similar between the LL and LT models with no obvious advantage for either model. The two models performed similarly with respect to accuracy and rank correlation of predicted genetic effects and recovered true values of genetic parameters and fixed effects, except for CE maternal heterosis from LL model. The effect of missing records was assessed using the LT model. All dispersion and location parameters were generally well recovered, even when the total proportion of missing records of both traits was up to 41%. Levels of missing CE and BW records that exist in Ontario do not seem to adversely affect genetic evaluation of CE. Key words: Accuracy, Gibbs sampling, heterosis, Snell score