Multiple Trait REML Estimates of Genetic Parameters for Holstein Cows for Milk Component Yields and their Log-transformations Using a Repeatability Model

Dairy Herd Milk Recording Association in Hokkaido. Years of calving were 1985 through 1991. Yield traits (2X, 305-d) used for the analyses were the fat, protein, and lactose (including minerals), and water yields. Data were divided into ten subsets of approximately 38, 000 cows each. The EM-REML method was used to estimate (co) variances with a multi-traitrepeatability model. The model contained herd-year, calving month, and the linear effects of ageat calving as fixed effects, and additive genetic, permanent environmental, and random residualas random effects. Estimates from ten data sets were pooled into the simple arithmetic mean. Variance components were estimated using a MTC program. Average heritabilities were .27, .28, .29, and .30 for protein, fat, lactose, and water yields, respectively. Heritabilities were increasedby .02 to .03 due to log-transformation, but repeatabilities were not increased by this amountbecause the fractions for the permanent environmental component were decreased by .02 in alltraits. Milk components except for fat yield were highly correlated with each other; in particular, the correlation between milk and water yields was 1.00. The log-transformation increased some of genetic correlations by as much as .03 and effectively eliminated the positive correlation between mean yield and variance component for additive genetic, permanent environmental or residual effects in subsets.