ARS | Publication request: ACCURACY AND STABILITY OF NATIONAL AND INTERNATIONAL SOMATIC CELL SCORE EVALUATIONS

Submitted to: Journal of Dairy Science
Publication Type: Abstract
Publication Acceptance Date: March 23, 2004
Publication Date: July 25, 2004
Citation: Powell, R.L., Sanders, A.H., Norman, H.D. 2004. Accuracy and stability of national and international somatic cell score evaluations [abstract]. Journal of Dairy Science. 87(Suppl. 1):87-88.

Technical Abstract: Most research on accuracy of dairy bull genetic evaluations has been with yield traits. This report focuses on somatic cell score (SCS) evaluations. International evaluations through Interbull (IB) have been available since May 2001. February 2004 U.S. and IB SCS PTA were matched with U.S. and IB PTA from May 2001 (U.S.04, IB04, U.S.01, and IB01, respectively) for 14,821 Holstein bulls. All four evaluations were required for each bull. Over the three yr, the means were essentially the same (increased 0.001) in each system (U.S. and IB). Correlations of 0.96 and regressions of later on earlier PTA of 0.99 were nearly as expected. Bulls adding substantially to numbers of U.S. daughters had mean increases in PTA of about 0.01 and regressions of nearly unity but correlations between earlier and later PTA were about 0.04 lower than expected considering reliabilities. The U.S.01 and IB01 PTA were compared as predictors of U.S.04 where IB01 PTA included data from at least one other country. As expected, U.S.01 was the better predictor where it contained essentially the same daughters as U.S.04, but IB01 was better when U.S. daughters increased at least 25%. Mean PTA difference from US04 was closer to zero, standard deviation of the differences was smaller, and correlations were higher. The advantage in correlation from IB01 increased with the amount of new U.S. data in the 2004 evaluation. Both U.S. and IB evaluations were stable on average, increasing in mean only slightly. Regressions were essentially unity in both systems. Correlations were less than expected by about 0.01 overall and lower by 0.04 for subsets with substantial added data. Inclusion of foreign data in addition to U.S. data improved the prediction of later U.S. SCS evaluations.