REPEATABILITY AND BIAS OF ESTIMATED BREEDING VALUES FOR DAIRY BULLS AND BULL DAMS CALCULATED FROM ANIMAL-MODEL EVALUATIONS

An animal model and an approximative method for calculating repeatabil ities of estimated breeding values are used in Finnish dairy cow evalu ation. Changes in estimated breeding values over time as daughters acc umulate were studied. Special emphasis was given to the accuracy and p otential bias in the pedigree indices of young sires. The data set use d was the same as in the national evaluation-and the traits investigat ed were protein yield and somatic cell count. The average repeatabilit y in evaluation of bulls without daughters was 0.37. The empirical rep eatability defined as a squared correlation between the pedigree index and the final sire proof was only 0.15. The reduction in the repeatab ility was attributed to the selection on pedigree index. The upward bi as observed in pedigree indices was 5 kg (approx. 0.3 of genetic stand ard deviation). The bias was caused by the overestimation of bull dams ' breeding value. Also the proofs of bull sires increased after the se cond crop of daughters. The correlation between the evaluations of the same sire calculated from two separate equal size daughter groups was 0.91 when the bull had 10 to 50 daughters and 0.87 with over 100 daug hters. This illustrates how the relative weight of the pedigree decrea ses while more progeny information is accumulated in the evaluation.