Curva de crescimento de codornas ajustadas por modelos não lineares

Abstract

This study aims to compare non-linear models fitted to the growth curves of quail in order to determine which model best describes the growth , to verify the similarity of the models by analyzing estimates of the parameters and to apply the model identity in different groups of strains and Treatments . We used the weight and age data from European quails (Coturnix coturnix coturnix) from three strains, in a 2x4 factorial design, housed in a completely randomized design with two levels of energy and four levels of crude protein, with six replications. Non-linear models were used: Brody, Von Bertalanffy, Richards, Logistic and Gompertz. To choose the best model we used the adjusted coefficient of determination, the percentage of convergence, the Mean Square Residue, the Durbin - Watson test, the Akaike information criteria and the Bayesian information criteria as evaluators of the quality of fit. We used cluster analysis to verify, based on average estimates of the parameters, the similarities between the models. We tested the identity of the models in order to verify whether the curve parameters for the eight treatments and for the three strains, would be equal. Among the models studied, the Logistic was the most appropriate to describe the growth curves. The Logistic and Richards models were similar in the analyzes irrespective of strain, so the analysis of Lineage 1, 2 and 3. There was no significant difference for the chosen parameters between treatments within each strain, however, there was significant difference, except for treatment 5, for the parameters of strains within each treatment, caused by the adult weight.