|O'CONNELL, JEFFREY - University Of Maryland School Of Medicine|
Submitted to: Interbull Annual Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 2/26/2018
Publication Date: 2/26/2018
Citation: Van Raden, P.M., O'Connell, J.R. 2018. Validating genomic reliabilities and gains from phenotypic updates. Interbull Bulletin. 53:22-26.
Interpretive Summary: Current validation of genomic evaluations of dairy bulls by the Interbull Centre (Uppsala, Sweden) tests estimates of genetic merit (breeding values) to ensure that they are unbiased and accurate. However, the validation tests use information not familiar to breeders, and the accuracy measure (reliability) published with evaluations is not tested. A method was developed to validate evaluation reliability based on the variability of differences between earlier and later estimated breeding values, which is familiar to breeders. This new method avoids downward bias that can result from selection. Published reliabilities were found to be slightly high for U.S. Holsteins and slightly low for U.S. Jerseys. Similar formulas based on reliability differences can estimate potential gains from more frequent phenotypic updates (such as monthly or weekly updates).
Technical Abstract: Reliability can be validated from the variance of the difference of earlier and later estimated breeding values as a fraction of the genetic variance. This new method avoids using squared correlations that can be biased downward by selection. Published genomic reliabilities of U.S. young bulls agreed closely with validation formulas. Similar formulas based on reliability differences can estimate potential gains from more frequent phenotypic updates such as monthly or weekly updates, but those gains are small when young animal reliabilities are high.