Correcting overestimation of approximate traditional reliabilities with herd-sire interactions when young genomic bulls are used in few herds

Authors: TABET, JOE-MENWER - University Of Georgia; AGUILAR, IGNACIO - Collaborator; BERMANN, MATIAS - University Of Georgia; LOURENCO, DANIELA - University Of Georgia; MISZTAL, IGNACY - University Of Georgia; Van Raden, Paul; VITEZICA, ZULMA - Collaborator; LEGARRA, ANDRES - University Of Georgia

Submitted to: Genetics Selection Evolution
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/16/2025
Publication Date: 6/23/2025
Citation: Tabet, J., Aguilar, I., Bermann, M., Lourenco, D., Misztal, I., Van Raden, P.M., Vitezica, Z.G., Legarra, A. 2025. Correcting overestimation of approximate traditional reliabilities with herd-sire interactions when young genomic bulls are used in few herds. Genetics Selection Evolution. 57:33. https://doi.org/10.1186/s12711-025-00984-0.
DOI: https://doi.org/10.1186/s12711-025-00984-0

Interpretive Summary: Each animal’s estimated genetic merit has a reliability to indicate its accuracy, but the reliability estimates previously ignored the effects of genetic-environmental interactions. Thus, reliabilities were over-estimated for bulls with many daughter records in a few herds when models included sire-by-herd interactions. Fast, approximate methods were developed to adjust for those effects in the national genetic evaluation. Those methods remove biases in the estimates and will improve farmer confidence that the reliabilities are estimated correctly regardless of the data pattern.

Technical Abstract: Background Recent changes in management practices have led to the occasional extensive use of some bulls in a limited number of herds, resulting in possible preferential treatment of their daughters (equivalently, increased environmental covariance between daughters of a sire within the same herd), which is modelled as herd-sire effect (HS). Although this effect may be accounted for in genetic evaluation models, standard approximation methods for reliabilities do not consider it correctly, leading to an overestimation of some sires’ reliabilities. This study assessed the potential bias of these reliability estimates due to the HS effect in both simulated and real dairy cattle records, comparing with the standard method by absorption of the effect (standard Tier-Meyer), a single-trait existing method (Misztal-Wiggans) using iteration through the pedigree, and a proposed one (modified Tier-Meyer) more suited to multiple trait and random regression models. Results We observed that in the presence of HS effects, reliabilities obtained by approximations were generally overestimated for sires with many daughters in a limited number of herds, indicating that the initial approximation methods by absorption in the standard Tier-Meyer did not correctly capture the HS effect. The results demonstrated that both the Misztal-Wiggans and modified Tier-Meyer methods performed well in simulated and real data, resulting in unbiased estimates of reliabilities and higher correlations between true and approximated reliabilities. For cows, the inclusion of the HS effect had minimal impact. Conclusions This study identified bias in reliability estimates of sires with daughters concentrated in a few herds due to the HS effect, even when there are large numbers of daughters. This bias occurs when the HS effect is not correctly modeled in reliability approximation methods. Methods that specifically accounted for the HS effect produced unbiased reliability estimates.