Prediction of voluntary dry matter intake in stall fed growing goats

• Autor(a) principal: Almeida, Amélia Katiane de [UNESP]
• Data de Publicação: 2019
• Outros Autores: Tedeschi, Luis Orlindo, de Resende, Kléber Tomás [UNESP], Biagioli, Bruno [UNESP], Cannas, Antonello, Teixeira, Izabelle Auxiliadora Molina de Almeida [UNESP]
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1016/j.livsci.2018.11.002; http://hdl.handle.net/11449/189885
• topic: DMI; Goats; Modeling; Monte Carlo Risk Assessment; Sensitivity analysis; Tornado plot

A Monte Carlo Risk Assessment (MCRA) was used to investigate the variability of existing empirical equations to predict dry matter intake (DMI) for weaned Saanen goats. Probability distribution functions were generated for each input variable used in the investigated DMI predictive equations using the Monte Carlo technique, and Spearman correlations (ρ) among the input variables were used to maintain their observed correlation. Probability distribution functions were obtained using an evaluation database containing 515 observations from four studies with Saanen goats (14.4–48.7 kg body weight (BW)). Thus, the pattern of the probability distribution functions relied exclusively on the observed distribution of the input variables. The MCRA simulation had 5000 iterations and used the Latin hypercube sampling approach to enable a balanced sampling throughout the distribution. Subsequently, with the Monte Carlo simulations, we generated tornado plots using standardized regression coefficients to evaluate influential input variables, and estimated the overlap between observed and predicted DMI. The overlap provided the percentage similarity considering the entire distribution shape. Additionally, each extant DMI equation was challenged by varying the input variables (i.e., independent variables) within the 90% confidence intervals of the probability distribution functions to obtain the prediction range of each equation. Finally, we regressed residual (observed – predicted) values on the predicted values centered on their mean values for each extant DMI equation to assess their mean biases. Our results indicated that even though it is clear that DMI is influenced by goat size (i.e., BW, BW0.75, metabolic weight (MW)), significant biases were observed in all tested equations. Six out of ten literature equations tested did not show a mean bias, whereas only one among the ten tested equations did not have a linear bias. Sex class influenced ADG, age, DM digestibility, metabolizability, and relative size (i.e., inputs considered in some tested equations), and DMI (i.e., male goats had 8% greater DMI per unit of BW than females). Tornado diagrams revealed that BW was the most influential input in the equations commonly used for estimating DMI. Thus, goat size (i.e., BW, BW0.66, MW) is a potential reliable predictor of DMI. Given its influence in predicting intake, the dietary NDF would be considered when developing empirical equations. Future studies should focus on defining the role of environment in DMI regulation, and determining an accurate way to adjust DMI considering metabolic regulation mechanisms in goats.