The description of the growth curve in domestic animals is of importance in management and economic decision-making. The aim here was to determine the best non-linear mixed model to adjust the growth curve in commercial turkeys. The data come from an intensive turkey farm under a subhumid tropical climate. The live weight records of 266 female and 275 male turkeys, weighed weekly, from birth to 23 weeks, were used. The models of Gompertz, yt = A × exp(-b × exp(-k × t)), and von Bertalanffy, yt = A × (1-b × exp(-k × t))3 were used to estimate parameters and predict the growth curve; where: yt = live weight at the t-th week of age; A = the expected mature weight; b = the integration constant; k = the maturation rate. Six non-linear models using the Gompertz, and von Bertalanffy functions: one with only fixed effects, four mixed models considering the fixed, 1 to 3 random effects, and a last model including the random effect of turkey were used. The analyses were performed using the NLMIXED procedure of SAS, and the selection of the best-fit model was chosen based on the Akaike (AIC) and Bayesian (BIC) information criteria. AIC and BIC values improved with the inclusion of 1 to 3 random effects, in both models for females and males. Based on AIC and BIC criteria, the best mixed NLM was the model that included random effects for A, b, and k. However, the predicted weight values of the mixed models were similar.
Read full abstract