Abstract

For dairy cattle evaluation systems that use an anim al model, algorithm s are needed that aim for a maximum reduction in computing time. D ifferent data sets m ay vary in convergence rate for solving equations. In West Germ any, herds generally are sm all, which allows block inversion of equations for a system structured by herds. Using this strategy, a rapid rate o f convergence was observed for data sets of 67,953 and 190,842 cows. If starting values w ere used, 35 rounds o f iteration w ere sufficient. Use of starting values was m ore im portant for larger data sets. Partial exclusion of pedigree inform ation had little im pact on convergence rate.
 

H. H Swalve, R. Vogel-Lackenberg, R. Reents, E. Bruns

Proceedings of the World Congress on Genetics Applied to Livestock Production, Volume XIII. Plenary lectures, molecular genetics and mapping, selection, prediction and estimation., , 378–381, 1990