Comparison of sire evaluation methods for carcass gain using field data

Four methods of predicting breeding values for carcass gain are compared: “regressed” least squares (RLSQ), mixed model (MM), mixed model with relationships among sires considered (MMR) and contemporary-comparison (CC). Very small changes in the ranking of the sires were observed from one method to another (rank correlations > 0.96). Consideration of the relationships among the sires or deleting the breed-group factor from the model did not lead to substantially different breeding values. The good correspondence of the predictors is mainly due to the well balanced distribution of the progeny over herds.