Published online by Cambridge University Press: 02 September 2010
Data from 22, 28, 36 and 41 pregnant cows in weeks 4, 3, 2 and 1 before calving respectively and 51, 65, 56 and 36 cows in weeks 1 to 4, 5 to 8, 9 to 12 and 13 to 16 of lactation respectively were used in multiple regression analyses to study the feasibility of predicting the daily total and the roughage organic matter and digestible organic matter intakes of cows given known weights of concentrates plus roughage ad libitum. The digestibility of the total intake and of the roughage component, the weight of concentrates eaten per day, the proportion of roughage in the total intake, age, milk yield, fat-corrected milk yield, live-weight gain and the 1·0, 0·73 and 0·66 exponents of live weight were used as independent variables.
The accuracy of the models was assessed from the significance of the partial regression coefficients, the size of the residual standard errors, and the population tolerance limits about predicted values, and by applying selected models to a set of data not included in the study. The results indicated that the precision likely to be achieved in the prediction of intake for individual cows was of limited practical application, but the mean intakes of groups of 30 similar cows could be predicted with an acceptable level of accuracy for practical application. In general, weight of concentrates offered and the digestibility of the roughage were the most effective independent variables in all cows, together with milk yield in lactating cows. The application of the results to conventional feeding systems and to complete diets offered ad libitum is discussed.