Accounting for herd, year and season effects in genetic evaluations of dairy cattle: a review

Contemporary groups, in genetic evaluations of dairy cattle commonly formed by fitting herd-year-season effects, can either be treated as fixed or random effects. For population structures that result in contemporary group subclass sizes of about 15 effective daughters or more, fitting contemporary group as a fixed effect to remove bias from genetic evaluations seems to be appropriate. For situations that result in contemporary group subclass sizes of about four effective daughters or less, the amount of information lost can be tremendous, and a compromise between accounting for bias and reducing prediction error variance should be found. Models that have been investigated to find such a compromise have in common that the amount of information lost is being reduced by either (i) fitting some HYS effect (or parts of it) as random, random and correlated, or random effects nested within a fixed effect or (ii) increasing the actual number of observations within subclasses by forming herd classes or by forming seasonal groups using clustering methods. The effectiveness of a model depends very much on the population structure, mainly on the average herd size, so no general statement about the appropriateness of a model can be made. The results of test-day (TD) models which, in contrast to 305-day yield models, allow a direct correction for effects specific to the time of recording of a cow's performance, clearly showed the importance of effects present at the day of test. Overall, it can be concluded that the definition of contemporary group effects is somewhat arbitrary and has often been a compromise between bias and the effective number of daughters. Therefore, methods, which allow a definition of cutoff points for contemporary groups only at points at which significant changes in environmental conditions actually occur, should be investigated. (C) 1997 Elsevier Science B.V.