Seleção e classificação multivariada de modelos não lineares para frangos de corte

Abstract

The aim of this study was to classify non-linear models used to describe the growth curve of broilers using the cluster analysis technique, taking into account the results of different measures of quality adjustment regression. For this purpose, we used data of body weight and age the following genetic groups of broilers: Cobb500, Hubbard Flex and Ross308, of both sexes, thus constituting six classes. Ten non-linear models were fitted, the quality of fit was measured by the adjusted coefficient of determination, Akaike information criteria and Bayesian, mean square error and index asymptotic. Cluster analysis indicated the Logistico, Michaelis Menten, Michaelis Menten Modificado and von Bertalanffy models as the most appropriate description of the growth curves for the six classes studied.