Growth curve of different strains of broiler chickens/Curva de crescimento de diferentes linhagens de frango de corte caipira

This study aimed to identify the nonlinear regression model more adequate to describe and compare the growth curve of four strains of broiler chickens. One hundred birds were used, 13 males and 12 females from each strains. There were tested 13 models of nonlinear regression and after that the best model was used to test the parameters equality and the models identity. The average weight at 77 days of age was 3.017, 2.781, 2.724 and 2.587 kg to the strains Hulking, Mixed, Carijo and Naked Neck, respectively. The Quadratic Logarithmic model showed better fit the data from Hulking, Carijo and Mixed strains, according to the mean square and mean absolute deviation tests. To the Naked Neck line the statistical criteria showed as more suitable the Inverse Polynomial model followed by the Quadratic Logarithmic. The males and females growth curves were different between and within the strains. The divergence between the evaluated classes was attributed to the decrease of weight rate, c parameter of Quadratic Logarithmic model. Quadratic Logarithmic model could be used to describe the growth curve of the four strains. The Hulking strain presented the faster and the Naked Neck strain presented the slowest growth. The feed system with the objective of change the growth curve must be adopted by the 35th day of age, to attend the current legislation and the consumer market. Key words: quadratic logarithmic, body weight, nonlinear regression. Objetivou-se identificar o modelo de regressao nao linear mais adequado para descrever a curva de crescimento de quatro linhagens de frangos caipira e compara-las entre as linhagens. Foram utilizadas 100 aves, sendo 13 machos e 12 femeas de cada uma das linhagens. Foram testados 13 modelos de regressao nao linear e, apos a selecao do melhor modelo, foi verificada a igualdade de parametros e identidade de modelos nao lineares. O peso medio aos 77 dias de idade foi de 3.017, 2.781, 2.724 e 2.587kg para a Pesadao, Mista, Carijo e Pescoco Pelado, respectivamente. O modelo Quadratico Logaritmico apresentou melhor ajuste dos dados para as linhagens Pesadao, Carijo e Mista, segundo o quadrado medio do residuo e desvio medio absoluto. Para a linhagem Pescoco Pelado, os criterios estatisticos apontaram como mais adequado o modelo Polinomial Inverso, seguido do Quadratico Logaritmico. As curvas de crescimento entre machos e entre femeas das linhagens foram diferentes, assim, como as curvas de crescimento de machos e femeas dentro de linhage