<b>Validation of models to predict milk urea nitrogen concentrations, using herd intake to estimate individual dry matter intake</b> - DOI: 10.4025/actascianimsci.v28i1.666

Abstract

The objective of this study was to validate prediction models for milk urea nitrogen (MUN), aiming at evaluating utilization efficiency of diets for dairy cows. Data (8,833 observations) from Holstein cows were used. Milk production, body weight, days in milk and lactation number were collected on the milk sampling days. Milk samples were collected monthly for composition analysis. In order to evaluate the three models, accuracy, precision and robustness were tested. Despite being more accurate (mean bias=-0.21 mg/dL), model 3 was less precise (residual error=5.24 mg/dL) than model 2 (mean bias=-1.72 and residual error=4.90 mg/dL), while model 1 was the least accurate (mean bias=2.93 mg/dL) and the least precise (residual error=5.99 mg/dL). There was lack of robustness for the three models. Models for predicting MUN differ among themselves with regard to accuracy, precision and robustness.