Skip to main content
ARS Home » Research » Publications at this Location » Publication #236740

Title: Best prediction of lactation yields accounting for regional and seasonal differences

item Cole, John
item Null, Daniel

Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/5/2009
Publication Date: 7/12/2009
Citation: Cole, J.B., Null, D.J. 2009. Best prediction of lactation yields accounting for regional and seasonal differences. Journal of Dairy Science. 92(E-Suppl. 1):209(abstr. T38).

Interpretive Summary:

Technical Abstract: In the United States, lactation yields are calculated using best prediction (BP), a method in which test day (TD) data are compared to breed- and parity-specific herd lactation curves that do not account for differences among regions of the country or seasons of calving. This may result in biased estimates of lactation yields. Data from 5,345,621 lactations of 348,123 Holstein cows with lactation lengths between 250 d and 500 d, records made in a single herd, at least five reported TD, and twice-daily milking were extracted from the national dairy database (NDDB). Herds were assigned to one of six regions of the country, individual lactations were assigned to three-month seasons of calving, and lactation curves for milk and fat yield were estimated by parity group (first versus later) for regions, seasons, and seasons within regions. The resulting curves were added to the BP software and tested against a validation dataset of 891,809 lactation records from 400,000 Holstein cows sampled at random from the NDDB. Mature equivalent (ME) milk and fat yields were calculated using the standard curves and the new curves. Differences between 305-d ME yields were calculated and tested for significance. Yields calculated using 50-d intervals from 50 to 250 DIM and using all TD to 500 DIM allowed comparisons of predictions for records in progress (RIP). Differences in ME milk ranged from 0 to 51 kg and were slightly larger for first- than later-parity cows, but were not significant (P > 0.05) in any case. Fat yields also did not differ significantly. Correlations of projected yields for 50-d intervals with yields using all TD data were similar across analyses. Differences among predictions averaged 57 kg lower for the new curves than for the standard curves using TD in the first 50 DIM, decreasing to 20 kg as TD from subsequent 50-d intervals were added. Complete lactation yields were similar for all curves, but projected yields for RIP were slightly more accurate when adjusted for regional and seasonal differences.