Utilização de modelos não lineares para a avaliação de curva de crescimento em ovinos

Abstract

In view of the great growth potential of sheep farming in Brazil and the need to improve production rates in the country, the objective of this work is to adjust weight data - age of sheep from the Capim Branco sector, crossbred animals, with blood level 5/ 8 Dorper, crossing with Santa Inês, to non-linear regression models and from the estimates of the model parameters and the quality evaluators, indicate the best model to describe the growth of sheep. The data provided were from four experiments, from 2016 to 2019. Weighings were performed irregularly, varying the amount and intervals of weighing and were analyzed by individual adjustments. For the comparison of the fit of the models and definition of the most adequate, the quality evaluators were used - coefficient of determination (R²), mean square of error (QME), percentage of convergence (%conv) and the biological interpretability of the parameters. The analyzes were run in RStudio. Individual adjustments were made for the 70 animals for each of the models, Brody, Von Bertalanffy, Gompertz and Logistic. For the parameter a and k, Logistic presented better results. All models showed good results for QME and R². In the representation in estimated curves, the Logistic model stabilized when the animals reached sexual maturity, and the other models overestimated these values. Therefore, it was determined that the non-linear Logistic model was the most appropriate, as it obtained the highest percentage of data convergence (87.14%), high R² value (0.97), low QME value (2.76 ), and adequate asymptotic value (48.09kg), not tending to overestimate adult weight.