Predicting ad libitum dry matter intake and yield of Holstein cows

Two data files, one from New Hampshire (n = 3308) and one from Georgia (n = 678), containing 4-wk or weekly means, respectively, of ad libitum dry matter intakes (DMI) and related variables were used to predict DMI and yields of 4% fat-corrected milk and milk protein in lactating Holstein cows. The DMI ranged from 5.9 to 30.4 kg/d, and milk yield ranged from 5.8 to 64.3 kg/d. Because of the lack of data from < 14 d in milk, prediction was not possible for the first 2 wk of lactation. Factors considered for inclusion in the DMI prediction model were parity number (1 or > or = 2), treatment with bovine somatotropin (bST), day of year, days in milk, minimum (nighttime) temperature-humidity index, body weight, 4% fat-corrected milk yield, milk protein yield, and corn silage and total silage percentages in forage dry matter. In separate models, the silage predictors were replaced with more specific descriptors of ration dry matter, including percentages of crude protein, fat (ether extract plus soaps of fatty acids), concentrate, acid detergent fiber or neutral detergent fiber, and forage acid detergent fiber or neutral detergent fiber. The square and sometimes natural logarithm of predictors were included in models, which then were subjected to a stepwise backward elimination option of a multiple regression procedure. Several useful equations were developed to predict ad libitum DMI; the best models accounted for about 80% of the variability in DMI, and standard deviations were < 9% of mean DMI. Depression in DMI related to heat stress was higher in pluriparous cows than in primiparous cows (22% vs. 6%). The negative coefficient for effects of bST treatment on DMI suggested that milk yield increased proportionally more in response to bST than did DMI. About 74 to 77% of DMI predictions were within 2 kg/d of observed DMI.