Breeding Value Estimation for Environmental Sensitivity on a Large Dairy Cattle Data Set

Animal robustness, or environmental sensitivity, may be studied through individual differences in residual variance. These differences appear to be heritable, and there is therefore a need to fit models having breeding values explaining differences in residual variance. The aim of this report is to study whether breeding value estimation for environmental sensitivity (vEBV) can be performed on a large dairy cattle data set having around 1.6 million records. Two traits were analyzed separately, somatic cell score and milk yield. Estimation of variance components, ordinary breeding values and vEBVs was performed using standard variance component estimation software (ASReml), applying the methodology for double hierarchical generalized linear models. Converged estimates were obtained by running ASReml iteratively 20 times, which took less than 10 days on a Linux server. The genetic coefficients of variation for environmental variance were 0.45 and 0.52, for somatic cell score and milk yield, respectively, which indicate a substantial genetic variance for environmental variance. This study shows that estimation of variance components, EBVs and vEBVs, is feasible for large dairy cattle data sets using standard variance component estimation software.