Evaluation of Equations Based on Animal Factors to Predict Intake of Lactating Holstein Cows

Abstract

The accuracy of seven DMI prediction equations based only on animal factors was evaluated with 11 independent data files. Mean square prediction error was used to compare equation accuracy, which was considered to be unsatisfactory when the square root of the mean square prediction error was greater than ±20% of the observed mean DMI. Robust intake equations that have a tolerable level of prediction errors for most data files would be less risky for practical use than models that are highly accurate for some data files but highly inaccurate for others. The number of independent data files for which equation accuracy was unsatisfactory was used to measure lack of robustness. No equation evaluated was able to predict individual cow DMI with a prediction error that was consistently lower than ±20% of the observed mean intake. The most robust equation in this study predicted intake unsatisfactorily for 3 of the 11 evaluation data files. Unsatisfactory accuracy for this equation was mainly due to mean bias.