Impact of Selection on Components of Variance and Heritabilities of Canadian Holstein Conformation Traits

Abstract
Sire and error variance components and heritabilities were estimated for 26 conformation traits using linear type classifications of 175,693 daughters of 6681 Holstein sires in 21,869 herd-round-classifier subclasses. Estimates from first classifications during first lactations were contrasted with estimates from the same 175,693 females' most recent classification, which included reclassifications. Only cows that have their final classification raised a full category receive an official reclassification record; thus, heritability estimates from data that include these reclassification records may be biased. A total of 9420 (5.4%) of the first lactation, first classifications were replaced by reclassifications for the second analysis, which is current practice for official sire evaluations for Holsteins in Canada. Records, expressed as Snell's scores, were preadjusted for age within parity and stage of lactation at classification. Sire and error components of variance were estimated by Henderson's new method. When reclassifications replaced first classifications in the analysis, estimates of sire and error variances increased 7.2 and 11.1% respectively, as compared with the first lactation, first classification estimates. Heritabilities ranged from .07 for rear heel to .38 for stature and decreased by an average of 2.5% when first lactation, first classifications were replaced by reclassifications. Results indicate that the inclusion of reclassifications tends to bias heritabilities downward.