Abstract
The efficiency of the mixed model procedures used for sire evaluation depends on the statistical model and how well the model describes the real situation. A model ignoring maternal grandsire of the daughter was compared to a model where the effect of maternal grandsire was included. The evaluation was made on Swedish dairy records of first lactation cows and comprised 590 and 226 thousand records respectively, of daughters of Swedish Red and White and Swedish Friesian AI-sires. A threefold increase in the ties was registered between sires, thus giving a more accurate comparison between generation of sires. The standard deviation of predicted differences increased, indicating an increase in the amount of information utilized. The correlation between the PDs of the two models was close to unity, 0.98–1.00. The difference between the PDs estimated with both models for the same sires was correlated to a calculated genetic difference between the mates of sires. The correlation was 0.5–0.8 for the youngest groups of sires, showing that the model operated as expected. Results indicating negative assortative matings were found among the selected sires, which could be due to genetic differences between sire lines or to the superior sires beeing used with older cows.
Key Words: