Statistical Methodologies for the Prediction of Dry Matter Intake in Beef Cattle Raised on Pasture

his study aimed to develop and evaluate models to predict total dry matter intake (DMI) and pasture dry matter intake (PDMI) in beef cattle raised under tropical conditions. Data from 38 studies conducted in Brazil between 2006 and 2019 were used, totaling 881 animals. Three model classes were analyzed: model 1 (animal effects), model 2 (animal + supplement), and model 3 (animal + supplement + pasture). Predictions were obtained using linear mixed models and Bayesian inference. Bayesian models, particularly model 3, showed higher precision and accuracy, with greater R² values. For PDMI, Bayesian model 1 was the most effective. For DMI, Bayesian model 3 performed best, with the interaction between supplement intake and pasture crude protein being significant. Key variables included initial body weight (IBW), average metabolic body weight (ABWmet), and average daily gain (ADG), especially the IBW × ADG interaction. Evaluation metrics (MSE, RMSE, CCC) indicated lower error and higher concordance in Bayesian models. The Model Selection Index (MSI) showed that mixed model 3 had the lowest MAPE and highest Cb. In conclusion, the Bayesian approach provides a significant advantage by incorporating multiple sources of variation and uncertainty, whereas mixed models may be more efficient in specific cases.