Estimates of variances due to sire × herd interactions and environmental covariances between paternal half-sibs for first lactation dairy production

Abstract

The concepts of a sire × herd interaction and of an environmental covariance, common to all daughters of the sire in a herd, are statistically equivalent. Estimates of the variance due to interaction reflect both interactions and common environmental effects and vice versa. Hence, environmental covariances were estimated as the variance due to a sire × herd (or herd-year-season) effect and c 2 effects were defined as the proportion of the phenotypic variance due to this interaction. Estimates were obtained for milk, fat and protein yield and fat and protein content. Data consisted of lactation records of 391 156 British Friesian-Holstein heifers, calving between November 1981 and October 1983. Analyses were carried out considering records from pedigree, non-pedigree and “mixed” herds, separately. Effects of transforming data to logarithmic scale and of standardizing records in each herd according to the pooled within-herd-year-season variance or respective coefficient of variation were investigated. Interaction components of variance amounted on average to half of the respective sire components. Estimates of c 2 were higher for yields (2.1–4.3%) than for contents (1.7–2.9%) and higher fitting a sire × herd-year-season than a sire × herd interaction. Values were considerably lower for non-pedigree than for other herds. Transforming or standardizing records reduced c 2 throughout, indicating that estimates may have been inflated by differences in variability between herds. Variances between sires and heritabilities were markedly higher for pedigree data, differences could only partially be explained by scale effects or differing selection intensities. Sire evaluation procedures should account for c 2 effects to minimize bias due to “preferential treatment” and to avoid overestimates of the accuracy of sire proofs, in particular for limited herd-use sires.