|Van Tassell, Curtis - Curt|
Submitted to: BARC Poster Day
Publication Type: Proceedings
Publication Acceptance Date: 2/1/1997
Publication Date: N/A
Citation: N/A Interpretive Summary:
Technical Abstract: Estimates of variance components are required by the Animal Improvement Programs Laboratory (AIPL) for calculation of national genetic evaluations of dairy cattle. However, because of computational limitations, parameters have never been estimated using the complete national data set of lactation records maintained at AIPL. Although Ayrshire data (1% of the national data set) have been used for development of evaluation procedures, the complete data or large subsets will be used for estimation of parameters for all U.S. dairy cattle breeds. Parameters will be estimated with a multitrait model for milk, fat, and protein yields. The estimation method (method R) is a relatively new procedure that allows analysis of large data sets. Method R is based on calculation of R-values as functions of covariances of predicted random effects using "complete" data and random subsets of data and variances of predictions using data subsets. All R-values will be 1.00 if the parameters are appropriate for the population. Multitrait analysis extends this concept to R-values that represent covariances. Several search methods for parameters also are being evaluated: a "bump" algorithm that scales parameters based on regression value, Newton's method, and downhill simplex.