|Goddard, M - UNIV NEW ENGLAND,NSW,AS|
Submitted to: Proceedings New Zealand Society of Animal Production
Publication Type: Proceedings
Publication Acceptance Date: June 6, 1996
Publication Date: N/A
Interpretive Summary: Milk yield of dairy cattle typically is estimated from periodic measurement of milk volumes and analysis of milk samples for fat and protein percentages. These measurements historically have been on a monthly interval, but the need to reduce the cost of milk recording has made less frequent measurements common. For genetic evaluations, test-day y yields have been combined into a lactation measure. However, greater precision in accounting for environmental influences is possible by analyzing test-day yields directly. Using test-day records of milk, fat and protein yields from herds in Victoria, Australia, a computationally feasible test-day model that accounted for genetic differences within and across lactations was developed for computing genetic evaluations. This system should make evaluations more stable and more accurate for dairy breeders by removing biases due to genetic differences in persistency and d rate of maturity. This system also will serve as the prototype for further research on possible implementation of a test-day model for U.S. genetic evaluations for yield traits.
Technical Abstract: In a two-step analysis, test-day effects were estimated within herd with adjustment for across-herd effects, and then the adjusted data were analyzed across herd. Genetic effects were defined for each of 10 months in milk within first and later lactations and for milk, fat and protein giving 60 traits. The rank of the genetic (co)variance matrix (G) was reduced to 6 such that G retained the information to evaluate the selection objective. A repeatability model allowed for multiple lactations with each lactation, conceptually, expressing all 60 traits, but missing observations for 30 or more. A canonical transformation was applied to create uncorrelated traits. Missing values were replaced by their expectations at each round. Because of the rank reduction, only 6 canonical traits were solved for. This system should make evaluations more stable by removing biases due to genetic differences in persistency and rate of maturity.