Author
 |
POWELL, REX |
|
NORMAN, H |
|
Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 5/20/2000 Publication Date: N/A Citation: N/A Interpretive Summary: The use of genetic correlations less than unity in international dairy bull evaluation began in 1995. The result is that bulls from other countries are estimated as being lower in genetic merit on a particular country's scale than by assuming a genetic correlation of unity (i.e. that traits are exactly the same in different countries). The objective is to improve accuracy of estimation but effectiveness in doing that had not been studied. International evaluations from data in 1995 were computed both assuming unity genetic correlations and using estimated correlations. They were compared in their effectiveness in predicting national evaluations in 1999 for 8 countries. There was essentially no difference in accuracy for bulls with daughters in the particular country and at least one other country in 1995. Estimated genetic correlations were marginally useful for bulls without daughters in the country in 1995 evaluations. Countries with the lowest correlations with other countries were not included in the international evaluations in 1995, but results for countries with moderately low correlations compared to those with higher ones did not suggest conclusions would be different had those countries been included. Thus, the use of estimated genetic correlations (i.e. considering traits as somewhat different across countries) did not appear to result in important improvements in accuracy of evaluation. Technical Abstract: International Holstein bull evaluations for Canada, Denmark, France, Germany, Italy, The Netherlands, Sweden, and the US were calculated both with and without genetic correlations considered. Evaluations were recalculated by the International Bull Evaluation Service with data available in February 1995 using current methodology with estimated genetic correlations between countries of less than unity and with correlations considered to be unity (0.995). Results of the two sets of evaluations for milk, fat, and protein yields for each country were compared to national evaluations in 1999. The 1999 national evaluations were assumed to be the best estimates of true genetic merit on a particular national scale. The examination was restricted to those bulls with at least twice the number of national daughters in the 1999 national as in 1995 international evaluations. The doubling restriction was needed to reduce the impact of the part-whole relationship due to earlier national data in common for international evaluations and later national evaluations. Failure to do so would have provided results that favor the procedure that discounted the foreign data through application of genetic correlations less than unity. Correlations and standard deviations of differences from later national evaluations showed no advantage to considering genetic correlations over ignoring them. Data for bulls without national data in the earlier international evaluations had no data in common, and the advantage from using estimated genetic correlations for yield was small. Thus, the use of estimated genetic correlations appears of marginal value at best for those countries. |
