番鸭体重生长非线性混合效应模型评价的研究

【目的】定量描述和估计本地番鸭体重生长的轨迹和参数。【方法】本研究对116只番鸭进行了21 weeks的饲养和观测;利用5种生长函数(Gompertz、Logistic、Von Bertalanffy、Richards和Brody)拟合了番鸭体重生长的非线性混合效应模型;分别计算了这些模型的8个生长参数(成熟体重、拐点日龄、拐点体重、初生体重、绝对生长率、相对生长率、绝对成熟率和相对成熟率);根据信息准则、误差方差的大小和生长参数估计值的准确度,比较了不同模型间的拟合和估计结果。【结果】与非线性固定效应模型比较,5种非线性混合效应模型拟合的番鸭体重生长方程和估计的8个生长参数,其准确度整体上提高了许多;全群番鸭体重生长Gompertz非线性混合效应模型,其信息准则最小,误差方差减少量达99.92%;雄、雌番鸭体重生长Richards非线性混合效应模型, 其信息准则最小,误差方差减少量分别是89.80%、91.81%。【结论】Gompertz和Richards非线性混合效应模型分别是评价全群和雄、雌番鸭体重生长的最优模型。

Study on Nonlinear Mixed Effect Model for Evaluation of Muscovy Body Weight for Growth

【Objective】 The aim of the study is to describe and estimate the trajectory and parameters of body weight growth of native Muscovy. 【Method】 One hundred and sixteen Muscovy ducks were fed and observed for 21 weeks, five body weight growth functions including Gompertz, Logistic, Von Bertalanffy, Richards, and Brody were estimated, nonlinear mixed effect models were fitted to data of body weight in Muscovy, and the 8 growth parameters including mature weight, Point of inflection (POI), POI weight, birth weight, absolute growth rate, relative growth rate, absolute maturing rate, and relative maturing rate were computed for these models in this study. The results of fitting and estimating with these different growth models were compared using information criteria,error variance, and accuracy of estimated growth parameters for evaluating the alternative models. 【Result】 The 5 nonlinear mixed effect models with 8 growth parameters each were fit for body weight of Muscovy ducks. Their accuracies had a much larger increase than these of the nonlinear fixed effect models. Among the 5 nonlinear mixed effect models for the body growth of female and male Muscovy, information criteria of Gompertz growth model were least, and reduction of residual error variance of this model was up to 99.92%. For the body growth of male or female Muscovy, information criteria of Richards growth model was least among the 5 nonlinear mixed effect models. Reduction of residual error variance of this model was up to 89.80% for male group, 91.81% for female group. 【Conclusion】 Gompertz,Richards nonlinear mixed effect models are best growth models for fitting the body growth of Muscovy,for male or female Muscovy, respectively.