Prediction of energy balance from milk traits of Holsteins in Japan.

We predicted the energy balance (EB) of Holstein cows in Japan from milk traits obtained by herd testing. Records covered 156 lactations of 102 cows. The number of artificial inseminations was highest, and the first conception rate was lowest in the low-EB group. Four prediction models were developed-for the whole lactation and for the early, middle, and late stages of lactation-with 20 variables, covering days in milk (DIM), milk yield, and milk composition traits. The actual and predicted EB means agreed well within DIM classes; the means of the residuals were smaller in the lactation stage models than in the all-lactation model, but the standard deviations (SDs) of the residuals were similar among models. After data reduction, the SDs of the residuals for 100 iterations were <1 throughout lactation in both types of models when n = 100. After model reduction, including the daily change of milk yield as a variable minimized the SDs of the residuals. Our equations for herd-level EB prediction have potential for use in genetic evaluation.