Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Detection and Adjustment of Abnormal Test-Day Yields

Authors
item Wiggans, George
item Vanraden, Paul
item Lake, Jill

Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 19, 2003
Publication Date: August 1, 2003
Citation: Wiggans, G.R., Van Raden, P.M., Philpot, J.C. 2003. DETECTION AND ADJUSTMENT OF ABNORMAL TEST-DAY YIELDS. Journal of Dairy Science. 86(8):2721-2724.

Interpretive Summary: A method to detect and to adjust abnormally low or high milk, fat, and protein yields on test-day (TD) was developed. The TD yields are compared to previous and subsequent yields and are restricted be between a floor and ceiling based on predicted yield. Lactation yields are then calculated from the restricted TD yields instead of excluding tests for cows that are reported as sick. Coding of sick cows is not uniform across herds or processing centers lactation so records are calculated more uniformly by not relying on the sick code. Observed TD yields were defined as abnormal if <60% or >150% of predicted TD yield. When applied to data since 1997, 1.8% of milk, 3.4% of fat, and 1.9% of protein TD yields were outside this range. To test the adjustments, Lactation records with one abnormal TD yield or more were matched with subsequent lactation records. Correlation between consecutive lactations increased from 0.692 to 0.693 for milk (561,063 lactation pairs), from 0.653 to 0.660 for fat (951,387 lactation pairs), and from 0.686 to 0.694 for protein (488,653 lactation pairs). This improved the correlation between consecutive lactation yields indicates that this method improves accuracy in calculating lactation records. The method is applied routinely to TD records of cows that calved since 1997.

Technical Abstract: A method to detect and to adjust abnormally low or high milk, fat, and protein yields on test-day (TD) was developed. TD yields are compared to previous and subsequent yields and are restricted to be between a floor and ceiling based on predicted yield. Lactation yields are then calculated from the restricted TD yields instead of excluding tests for cows reported as sick. Coding of sick cows is not uniform across herds or processing centers so uniformity of calculation can be improved by not relying on the sick code. Observed TD yields were defined as abnormal if it is <60% or >150% of predicted TD yield. When applied to data since 1997, 1.8% of milk, 3.4% of fat, and 1.9% of protein TD yields were outside this range. Predicted TD yield was calculated as preceding yield plus preceding test interval multiplied by a daily yield change (slope) based on days in milk (DIM), DIM2, previous normal yield, and an interaction between DIM and previous yield. To accommodate changes in slope at peak yield, separate coefficients were estimated for <50 and >=50 DIM. To be designated as abnormal, TD yields that were abnormal based on previous normal TD yield also were checked against predicted TD yield based on previous and subsequent normal TD yields. Herd mean was used when fewer than three TD were recorded for a cow and to determine an acceptable range for component percentages. Predicted TD yield for first TD was based on subsequent rather than previous normal TD. To test the adjustments, Lactation records with one abnormal TD yield or more were matched with subsequent lactation records. Correlation between consecutive lactations increased from 0.692 to 0.693 for milk (561,063 lactation pairs), from 0.653 to 0.660 for fat (951,387 lactation pairs), and from 0.686 to 0.694 for protein (488,653 lactation pairs). The outlier adjustments improved the correlation between consecutive lactation yields and are applied routinely to TD records of cows that calved since 1997.

Last Modified: 11/25/2014
Footer Content Back to Top of Page