Comparison of mathematical models describing the growth of tropically adapted Ross 308 commercial broiler chickens

Abstract

Mathematical growth models are useful in describing the growth of livestock. The study was done to assess the predictive ability and accuracy of four three-parameter nonlinear mathematical models (namely: Gompertz, Gompertz-Laird, Logistic, and von Bertalanffy) and one four-parameter (namely: Richards) nonlinear mathematical model. Models were used to predict the body weight (BW) of commercial Ross broiler chickens adapted to tropical conditions (n = 1,286). Age-weight data were collected once every week for 6 weeks. The Gauss-Newton iterative process of the nonlinear procedure in SAS was employed to obtain the parameters for each model. In addition, each model's goodnessof-fit, residuals, and computational difficulty were estimated. Model parameters were evaluated using Akaike’s information criterion (AIC), Bayesian information criterion (BIC), adjusted coefficient of determination (AdjR2 ) and root mean square error (RMSE). The AdjR2 value for all five models was high; however, the highest value was observed in the Gompertz and Gompertz-Laird models. Furthermore, the lowest AIC, BIC and RMSE values were observed in the Gompertz models. Using a complimentary method (involving a subjective pairwise comparison of the observed and predicted BWs), the Logistic, Gompertz-Laird, von Bertalanffy, and Richards models fitted well for the data used. However, the best fitting was obtained in the Gompertz model. Some similarities were observed between the Logistic and Richards models. In conclusion, all five nonlinear mathematical models fitted the age weight data used in this study well, with the Gompertz model being the best.