Estimation of Breeding Value for Milk Production Traits in a Holstein Herd with Incomplete Relationships

The estimation of breeding value will be biased if it is estimated by animal model with incomplete information on the relationships. We investigated the bias of genetic parameters when relationships were not complete in sires. The 527 records of milk yield, fat yield, fat rate, protein yield and protein rate in 223 Holstein cows of Hokkaido National Agricultural Experiment Station were used. Relationships of cows with milk records trace back three generations and relationships of their sires trace back to their sire. Relationships were made up from 295 cows and 110 sires. Heritabilities, genetic correlations, year-parity effect and breeding values were estimated with restricted maximum likelihood in multitrait animal model for milk yield, fat yield, fat rate, protein yield and protein rate. The model contained year-parity and calving month of fixed effects and additive genetic random effect. The effect of age at calving for milk yield was adjusted before estimation of genetic parameters. Year-parity effect was distributed every year in equation (1) and every three years in equation (2). Heritability estimates were from 0.36 to 0.54 for milk yield, fat yield and protein yield and from 0.80 to 0.83 for fat rate and protein rate. These values from equation (1) are slightly higher than those from equation (2). The increase in year-parity effect with calving year by equation (1) was higher than that by equation (2) especially for milk yield, fat yield and protein yield. The breeding values of sires for milk yield, fat yield and protein yield were underestimated in equation (1) when they were born after 1980. This implies that genetic trend in sires could not be accounted for in the analysis of equation (1) due to incomplete relationship among sires. However, even in equation (1), breeding value of sires born after 1980 were not biased for the traits of fat rate and protein rate. This is attributed to the stable genetic trend for these traits. Breeding values of every five traits were not biased in equation (2). Large number of cows in the group of year-parity effect would result in more proper sire estimation especially when relationship is not enough in such a manner as represented in equation (2).