Wu Xiao-Lin, Cole John B, Legarra Andres, Parker Gaddis Kristen L, Dürr João W
Council on Dairy Cattle Breeding, Bowie, MD 20716.
Department of Animal and Dairy Sciences, University of Wisconsin, Madison, WI 53706.
JDS Commun. 2025 Jun 9;6(5):675-680. doi: 10.3168/jdsc.2024-0668. eCollection 2025 Sep.
Accurate genetic evaluations rely on high-quality phenotypic data; however, measurement errors and data inconsistencies-such as those arising from unsupervised or incomplete sources-pose challenges to their reliability. This study investigates the effect of response errors on genetic evaluations across continuous and categorical traits. We introduce an additive measurement error model to illustrate how phenotypic errors influence genetic effects and variance estimation. Next, we examine a binary trait scenario, demonstrating the utility of sensitivity and specificity in adjusting observed incidence rates for misclassified data. To further illustrate genetic evaluation in the presence of misclassifications, we proposed a mixed effects liability model assuming unequal sensitivity and specificity or varied false-positive and false-negative rates. Our findings underscore the necessity of integrating measurement error models into genetic evaluation frameworks to reduce bias and enhance predictive accuracy.
准确的基因评估依赖于高质量的表型数据；然而，测量误差和数据不一致性——例如那些源于无监督或不完整来源的误差——对其可靠性构成了挑战。本研究调查了响应误差对连续性状和分类性状基因评估的影响。我们引入了一个加性测量误差模型，以说明表型误差如何影响基因效应和方差估计。接下来，我们研究了二元性状的情况，展示了敏感性和特异性在调整误分类数据的观察发病率方面的效用。为了进一步说明存在错误分类时的基因评估，我们提出了一个混合效应易感性模型，该模型假设敏感性和特异性不相等或假阳性和假阴性率不同。我们的研究结果强调了将测量误差模型纳入基因评估框架以减少偏差并提高预测准确性的必要性。
