Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: June 22, 2003
Publication Date: N/A
Evaluations of bull dams may be biased upward because of preferential treatment or contemporaries with extremely low yields. Variance of Mendelian sampling (MS) was standardized to determine if such an adjustment of lactation records could improve accuracy of estimated breeding values (EBV) of bull dams. For Holstein data included in February 2003 USDA evaluations, MS variances were calculated within herd and 5 yr of first calving group. To regress estimates for small herds, the population estimate was included with a weight of 20. The ratio of the population MS standard deviation to the within herd-5 yr group value was used to adjust phenotypic yields. This ratio was limited to a maximum range of 0.5 to 2.0, with the range further limited based on mean herd yield to avoid over adjustment. To minimize effect of unreliable evaluations of females, MS was calculated using maternal grandsire EBV instead of dam EBV. At each round of iteration, this male index was subtracted from current EBV and the result adjusted for amount of information. To assess effect of the MS variance adjustment, EBV were calculated excluding records from calvings after 1997. Means of parent EBV were calculated for recent bulls using subset EBV with and without MS variance adjustment. As shown in the table, correlations between February 2003 bull EBV and parent mean for the subset were higher with adjustment. Similarly, regression (slope) of EBV on parent mean was closer to 1 with adjustment, and mean EBV minus parent mean (bias) was lower. Mean EBV of the top 100 cows was reduced with adjustment. Twenty percent in the top 50,000 cows for milk yield were different with adjustment. The largest drop was 1324 Kg EBV. Standardization of MS variance improved accuracy of EBV.