Breeding value estimation for somatic cell score in South African dairy cattle

Abstract

Two fixed regression testday models were applied for variance component estimation and prediction of breeding values for somatic cell score, using testday records of the first three lactations of South African Holstein and Jersey cows. The first model (ML-model) considered the testdays of the different lactations as different traits in a multiple-trait animal model and the second analysis (RM-model) treated later lactation records as repeated measures of the first lactation. Heritabilities from the RM-model were more in the range of literature estimates compared to that of the ML-model, i.e. 0.19 + 0.003 for the Holstein breed and 0.18 + 0.003 for the Jersey breed. Rank correlations indicated that minor changes occur in the ranking of proven sires between breeding values obtained from the ML- and RM-models. Although genetic correlations between parities are not unity, the RM-model estimates more competitive variances and requires extensively less computer time to predict breeding values compared to the ML-model and are therefore recommended for breeding value estimation on a national basis.


South African Journal of Animal Science Supp 2 2004: 32-34