Models for adjusting for herd, year and season effects in dairy sire evaluation in small herds

Abstract

Alternative models were examined for adjustments of herd, year and season effects for evaluation of dairy sires in Switzerland where average herd-year subclass size for first lactation records is less than 4. Herds are presently grouped into herd classes according to production level, and herd classyear-season is fitted as a fixed effect like herd-year-season in situations with large herds. Several models were compared fitting herd class vs. herd effect, accounting for the interaction with year and season as random or fixed effects. The empirical and predicted error variances of sire effects were estimated from 10 random subsets of the data of proven sires. The empirical and predicted error variances were consistently smaller with herd class models than with herd effect models. However the plots of sire effects showed that the herd class models overestimated poor sires and underestimated good sires. This bias was mainly because the herds with good management and poor sires were grouped in the same herd class as those with poor management and good sires. A herd class model with herd classification based on herd averages adjusted for genetic merit of sires showed no bias. A model with herds as fixed and herd-years as random effects was found to be a good alternative to a herd class model. Models accounting for herd-year × season interactions showed no advantage in the accuracy of evaluation.