Update on the TestDay Model

1
Update on theTest-Day Model
2
Test-Day Model

• Data includes individual cow milk weights and
  component percentages recorded on test day
• Genetic evaluations based on test-day (instead of
  lactation) data can account for environmental
  influences more accurately and for genetic
  differences in shape of the lactation curve

3
Test-Day Model continued

• Used in Canada and several European countries
  Australia has used a test-day model since 1985
• Cornell Foundation granted patent in 1993, which
  has prevented the U.S. from implementing a
  test-day model 2000 re-examination upheld the
  patent

4
Test-Day Model continued

• Canadian Dairy Network requested re-examination
  of patent in Canada expects to learn if patent
  was upheld during 2002
• Several European groups are collecting
  information to show that patent should not be
  upheld in Europe

5
Test-Day Data

• Received at least weekly from regional computing
  centers
• Includes information from nearly 30 million
  lactations from calvings since 1990
• Many checks to ensure data integrity

6
Test-Day Data continued

• Lactation records calculated using best
  prediction
• Upper and lower limits imposed
• 60-150 of expected yield derived from preceding
  or adjoining normal test days
• Floor and ceiling used
• 1-2 of test-day yields affected
• Test-days weighted according to correlation with
  305-day yield
• Weights affected by closeness to other test days

7
Test-Day Data continued

• Recent changes
• Minimum days in milk decreased to 5
• Specific reporting of lactations that start with
  an abortion or reasons other than a normal calving

8
Detecting Outlier Lactation Records

• Some herds have high percentage of "elite" cows
• Investigation showed abnormal distributions for
  lactations from those herds

9
Outliers continued

• Possible solution
• Develop criteria based on normality statistics to
  detect herds with outlier lactation records
• Adjust weight that records receive based on
  departure from normality
• For extreme cases, cow's own records would have
  little influence on her evaluation
• Determine if reduced weight for extreme records
  improves accuracy of parent average in predicting
  future progeny performance