Variance Reduction and Measurement Errors in Estimating Lactation Milk Yields Using Best Prediction: an Analytical Review

Best Prediction has been utilized in the United States to estimate unobserved daily and lactation yields from known test-day yields since 1999. This method has proven more accurate than its predecessors. However, it has 2 significant challenges in practice. First, Best Prediction reduces the variance of estimated yields compared with actual yields. Reduced phenotypic variance represents a concern because it can significantly underestimate genetic variations in genetic evaluations. Second, measurement errors occur in the projected lactation yields from incomplete or inaccurate test-day records. These errors can adversely affect the accuracy of lactation yield estimations and the subsequent genetic evaluations. This paper provides an analytical review of Best Prediction, focusing on variance reduction and measurement errors. We demonstrate how variance reduction and measurement errors can be intrinsic to the method. Illustrative examples are presented, highlighting the practical challenges and possible solutions.