DEVELOPMENT OF A MODEL FOR ESTIMATION OF BREEDING VALUES FOR BEEF-CATTLE IN GERMANY .1. MODELING THE HERD EFFECT

The analysis of modelling the herd effect for estimation of breeding v alues in beef cattle was based on field data of Charolais. Traits of i nterest were birth weight, weaning weight and yearling weight Models c onsidering herd and herd-year effects as fixed and random as well as m odels with a random herd-year-season effect were compared with each ot her. Criteria for the comparison were log-likelihood, residual varianc e, accuracy of prediction of breeding values and correlation of breedi ng values of sires based on two randomly divided data sets. Hierarchic al model comparison showed significance of the interaction between her d and year for the fixed as well as for the random effect model as ind icated by the highest log-likelihoods. When the interaction between he rd and year were additionally considered in the fixed effect model the accuracy of estimation of breeding values decreased from 0.67 to 0.63 (0.67 to 0.64) and the correlation between breeding values of two ran domly divided data sets decreased from 0.55 to 0.25 (0.52 to 0.36) for weaning weight (yearling weight). In contrast, when the interaction b etween herd and year were additionally fitted in the random effect mod el, the accuracy of estimation of breeding values increased from 0.67 to 0.72 (0.68 to 0.71) and the correlation between breeding values of two randomly divided data sets increased from 0.56 to 0.62 (0.56 to 0. 67) for weaning weight (yearling weight). These indicated the desirabi lity of the random model, but the concern was discussed that this mode l may result in biased estimation of breeding values due to non-random distribution of genetic potential over herd-years. However, a conclus ion based on weighing increased In accuracy and correlation between br eeding values with the possible biased estimation of breeding values a s well as the indication of random distribution of genetic potential o ver all herds; resulted in the recommendation of using the interaction of herd-year effects as random.