Logistic nonlinear mixed effects model for estimating growth parameters

Abstract

This study was undertaken to apply the logistic model with nonlinear mixed effects to model growth in Japanese quail. Nonlinear mixed models (NLMM) allow for the inclusion of random factors in a linear manner, which accounts for the between-individual variability and heterogeneous variance. A fixed effects model (M1) was compared with NLMM containing either 1 (M2) or 2 (M3) random effects using the residual error variance, −2 log-likelihood, Akaike information criterion, and Bayesian information criterion as the criteria for evaluating these alternative models. In M2, the between-bird variability was modeled by varying the asymptotic BW, which led to a 57% reduction in the residual variance compared with M1 in males. In M3, the between-bird variation was partitioned into variances due to varying asymptotic BW and the age at the inflection point. The residual variance in M3 was reduced by about 72 and 38% compared with M1 and M2, respectively, in males. The correlation coefficient between the actual and predicted BW for M1, M2, and M3 were 0.9887, 0.9955, and 0.9975, respectively. Similar results were found in females. The model evaluation criteria indicated that the mixed effect models fitted the data better than the fixed effect model because they account for between-bird variation. The use of NLMM is recommended for modeling growth data in poultry because the predicted BW at different ages is more accurate than using the mean prediction function of the fixed effect model.