Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 23, 1998
Publication Date: N/A
Interpretive Summary: To calculate genetic evaluations, lactation yields must be estimated from recorded milk weights and component samples. Many of the new milk- recording plans for U.S. dairy cattle differ widely in the number of milk weights recorded and the number of component samples taken. The current procedure used to estimate lactation milk yield is the test interval method. However, a recently develop procedure, the best prediction method, calculates the accuracy of records by considering the number and distribution of tests. The best prediction method also provides an estimate of lactation milk and component yields from test day data. This study investigated whether the test interval method or the best prediction method was more accurate for estimation of lactation yield. For traditional test plans, a moderate increase in accuracy (4%) was found for the best prediction method. For a.m.-p.m. and trimonthly testing, the improvement was 6 and 10%, respectively. Because 60% of the milk-recorded cows in the U.S. are enrolled in a.m.-p.m. test plans, the best prediction method is a viable alternative to the test interval method and should be more effective than the test interval method in accurately predicting actual lactation yields, thereby increasing the accuracy of management reports for dairy producers as well as the accuracy of genetic evaluations on which producers base breeding decisions.
A method with best prediction properties that condenses yield from all test days into measures of lactation yield and persistency has been proposed as a possible replacement for the test interval method and projection factors. The proposed method uses previously established correlations between individual test days. Milk weights representative of monthly, a.m.-p.m., and trimonthly test plans were examined to compare the accuracy of the best prediction and test interval methods for estimation of lactation yield. Individual milk weights or daily yields of 658 Canadian cows in 17 herds were selected to correspond to test intervals for 100,000 U.S. cows. For estimation from a.m.-p.m. records, the initial milk weight credited was selected randomly from the morning or evening milking and alternated thereafter. Trimonthly credits were from one of the first three designated test day weights selected randomly and from each third designated test thereafter. Correlations between actual and 305-day lactation yields estimated by the test interval method were .97, .96, and .93 for monthly, a.m.-p.m., and trimonthly testing and .97, .97, and .93 for the best prediction method. Standard deviations of the difference between estimated and actual yields for monthly, a.m.-p.m., and trimonthly testing showed that the best prediction method had 4, 6, and 10% greater accuracy than the test interval method. The advantage of the best prediction method is moderate if two milk weights are recorded monthly and greater if testing is less frequent. Estimation of fat and protein yields was more accurate with a multitrait best prediction method than with the test interval method for records with reduced component sampling.