Modeling differences in feed intake and efficiency: Growing and finishing beef cattle

Objective: Our objective was to evaluate model structure choice on estimates of efficiency, based on residual feed intake, and thereby cattle profitability. Materials and Methods: Efficiencies were compared for 2 models using data from 7 studies (509 cattle). Model 1 DMI were estimated using ordinary least squares as f(ADG and BW^0.750). Model 2 was a multivariable Bayesian model; DMI was f(ADG, BW^0.750 and BW). For model 2, DMI, BW, and ADG were ranked by model 1 residuals categorized in quintile to evaluate whether information found in residuals related to composition of gain or maintenance. Results and Discussion: Efficiency rankings lacked concordance and predictive value between models, suggesting that ordinary least squares, Bayesian, or both frameworks lack utility to predict efficiency. With the exception of 1 data set, DMI was better predicted in the Bayesian framework. Estimated recovered energy (Mcal/d) in model 2 was less for cattle in quintile 1 than in quintile 5 for 4 of 7 data sets and numerically less for 6 of 7 data sets. Estimated maintenance in model 2 was less (quintiles 1 vs. 5) for 5 of 7 data sets. Substantial information existed in model 1 residuals regarding differences in composition of gain and maintenance not found in model 2 residuals. Implications and Applications: Differing efficiencies between models indicate that residuals are properties of models, not cattle. Selection of cattle with less empty body fat may not be desirable from an economic standpoint.