Females in MACE: Resolving double counting using the reverse reliability approximation

Abstract

Including females’ estimated breeding values (EBVs) in Multiple Across-Country Evaluation (MACE) requires a way to prevent information shared through relatives from being counted more than once. This paper evaluates a reverse reliability approximation that converts submitted reliability into an adjusted effective daughter contribution (EDC) after accounting for pedigree information. The results focus on five Ayrshire populations contributing with female data: Australia, Canada, Great Britain, the United States, and South Africa. For cows, mean adjusted EDCs and reliabilities were compared with submitted national values for milk, fat, and protein. Adjusted mean EDCs were equal to or lower than national means in almost all population and trait combinations, with the largest reductions observed for Great Britain. Mean female MACE (FMACE) reliabilities were generally close to, but often lower than, national reliabilities; South Africa was the exception, with higher FMACE mean reliability for all three traits. Bull comparisons between conventional MACE and FMACE showed that adding female information increased international reliability across the participating populations. These results support the use of the reverse reliability approach as a practical mechanism for controlling double-counting when female information is incorporated into MACE. The results also show why EDC is central to the procedure: it represents the amount of information retained for an animal after reliability attributable through relatives has been accounted for.