MODELOS NÃO LINEARES EM DADOS DE CRESCIMENTO DE BOVINOS DE CORTE

Abstract

This study aimed to compare the performance of five nonlinear models—Gompertz, Meloun I, Meloun II, von Bertalanffy, and Weibull—in describing the growth of beef cattle under a confinement system. Data from 6,365 animals were analyzed, with weights collected at different productive stages, adjusted using nonlinear regression, and evaluated according to the following criteria: convergence percentage, coefficient of determination (R² and adjusted R²), mean prediction error (MPE), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC).The results indicated a trend of reduced weight variability throughout the productive cycle, especially during confinement, promoting uniformity within the herd. Among the models tested, the Gompertz model showed balanced performance (R² = 0.9208; convergence = 85.55%; MPE = 0.5619; AIC = 58.3891; BIC = 78.1959). The von Bertalanffy model achieved a higher R² (0.9421) and lower MPE (0.3563) but lower convergence (78.90%). The Weibull model stood out for the lowest AIC (55.8660) and BIC (76.1417) values, although limited by its low R² (0.6021). The Meloun I and II models showed unsatisfactory performance, with low explanatory power and high prediction errors.These results reinforce the applicability and relevance of the Gompertz model as a tool for nutritional, reproductive, economic, and genetic improvement planning in beef cattle production.