Skip to main content
ARS Home » Research » Publications at this Location » Publication #67939
Title: MATHEMATICAL REPRESENTATION OF RELATIONSHIPS BETWEEN DAILY MILK YIELDS

Author
item Norman, H
item Vanraden, Paul
item Wright, Janice
item MEINERT, TODD - NDHIA COLUMBUS, OH

Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/25/1996
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Daily milk yields for 650 Canadian lactations were used to derive correlations within herd-year between daily yields and between daily and 305-day yields. Smoothed functions were fit to remove sampling noise from estimated correlations and to minimize memory needed to predict 305-day yield. Correlations within herd-year between daily yields after the first 15 days in milk (DIM) were consistent and could be predicted with R2 of .88 from mean DIM and difference in DIM between two daily yields (linear, quadratic, and interaction). Although all 5 variables were significant, differences in DIM alone accounted for R2 of .86. Correlations were consistent across test days except for 1) daily yields early in lactation, which were low, and 2) daily yields late in lactation for which correlations declined at a slightly faster rate. Excluding daily yields during the first 20 or 30 days of lactation increased R2 to .91. Again, all 5 variables were significant, but interaction between DIM and difference in DIM accounted for an R2 of .89 and adding difference in DIM increased R2 to .91. Regression equations were developed to predict the correlation and, thus, covariance, which was then available to estimate 305-day yield from any possible combination of daily milk yields. Correlation between daily yield for a designated number of days was nearly always highest for mid-lactation, intermediate for late lactation, and lowest for early lactation. The 305x305 matrix of correlations between daily milk yields was described well by a simple function of mean and difference in DIM. Similar correlations for fat and protein yields will allow better prediction of missing component tests.