Modelos mistos em regressão não linear para descrição do crescimento de bovinos da raça nelore

Abstract

Birth body weights were used, 120, 210, 365, 450 and 550 days of age, referring to 2888 and 3348 male animals female animals Nellore created in the Central West of Brazil, aiming to describe the weight growth curve. Body weight of the animals was regressed by age (in days), by gender, using a nonlinear function Brody, Gompertz, Logistic, Meloun 1, Von Bertalanffy and modified Von Bertalanffy. The quality of model adjustment was verified by the asymptotic index, adjusted determination coefficient, Akaike information criterion and the Bayesian information criterion of Schwarz. Aside from the model that used the Richards function, all other models converged, with the model employed Brody function getting the best settings by evaluators. Using Brody function, but considering the asymptotic weight and maturation rate, parameters A and K respectively as random, a reduction in the residual variance estimated at approximately 79% and 83% for males and females, respectively, compared with the estimates obtained by these fixed context models. Males and females showed similar absolute growth rates in the early growth trajectory with 0.921 and 0.922 kg / day, respectively. With advancing age males had higher daily gains compared to females, 0.261 and 0.198 kg / day, respectively. An opposite behavior was observed for the relative growth rate, proportional gain mass to 2.55% and 2.39% for females and males, respectively, at the beginning of the growth trajectory and estimates equal to 0 and 0.06, 08% for females and males, respectively, at 550 days of age. Males showed greater acceleration of growth compared to females in the early growth trajectory, equaling up to females to 397 days of age and from that age had lower estimates of acceleration and growth. The regression model approach nonlinear mixed model in the context allows reduction in the residual variance, increase the accuracy of the model