NCR Model Evaluation for prediction of dry matter intake by dairy cows managed in tropical pastures

Abstract

The aim of the present study was to evaluate the predictive power of estimating the dry matter intake (DMI) of crossbred cows on tropical pastures by the National Research Council (NRC) equation for dairy cattle. A joint analysis of five studies covering three forages was performed in which 132 individual estimates of observed DMI obtained through Cr2O3 as a marker. DMI was also predicted from the NRC (DMIPred) software with inputs concerning animals and breeding environment of the studies used. Predicted DMIPred average values (12.7±1.6kg/d) were similar to the observed DMIObs ones (12.3±3.3kg/d). We obtained the following estimates of the evaluation of the predictive power of the model: average bias (-0.419kg/d), coefficient of determination (0.029), Person's correlation coefficient (0.17, P= 0.051), mean square error of prediction (11,844±20,034), model efficiency factor (-0.081), coefficient of determination (4.1032), and bias correction factor (0.767). The comparison between DMIObs and DMIPred values allowed the identification of the overestimating tendency of the predictions demonstrated by the robust regression fit of the simple linear no intercept model. Nevertheless, the model yields predictions with satisfactory accuracy, but with low precision.