Estimation of (co)variance components for the test-day milk yield in siboney cattle using a random regression model

In order to estimate (co)variance components for test-day milk in Siboney (518 Holstein 318 Cuban Zebu) cattle a random regression model was used and it was determined whether a heterogeneous residual variance (sigma(2)(r)) adjustment was necessary. Data were test-day yields of 17034 first lactations of 2086 Siboney cows, collected from 1995 through 2003. The cows were daughters of 153 sires and 1179 dams of 50 herds. The genealogy file was comprised by 5471 animals. It was guaranteed that the cows had, at least, the first four test-day milk controls. Random regression models were used testing Legendre polynomials of different order. The 4(th) order was selected for the additive variance and the 5(th) for the permanent effects (AG4PE5), based on the logarithm of convergence and the Akaike criterion. The sigma(2)(r) heterogeneity was applied to the model. The estimated heritability varied between 0.21 and 0.31 with maximum at 177 days of lactation. The genetic correlations were elevated and reached values >0.95 at mild lactation. The results emphasize the importance of modeling the sigma(2)(r) heterogeneity, because a better adjustment of (co)variance components for milk yield is obtained.