分类重复测量数据的非线性混合效应模型

THE NONLINEAR MIXED-EFFECT MODELS FOR CATEGORICAL REPEATED MEASUREMENT DATA

[目的]探讨分类重复测量数据的非线性混合效应模型及SAS8.0软件NLMIXED过程实现。[方法]直接拟合分类反应变量的非线性概率模型,结合重复测量资料的特点,采用附加高斯积分来获得最大似然的参数估计。[结果]非线性混合效应模型能很好地拟合分类反应变量的重复测量资料,它允许固定效应和随机效应进入模型的非线性部分,可方便地分析随机缺失等非均衡数据。[结论]分类反应变量重复测量资料的非线性混合效应模型分析结果合理、容易解释,为分类重复测量资料提供一种新的分析思路。

[Objective] To discuss the nonlinear mixed-effect models for categorical repeated measurement data and the procedure of realization in SAS8.0 software NLMIXED program. [Methods] The nonlinear probability models for categorical response variable were directly fitted and then combined with the characteristics of repeated measurement data, the maximum likelihood estimations were got by using adaptive Gaussian integration. [ Results] Nonlinear mixed-effect models could well fit the categorical repeated measurement data, both fixed and random effects were permitted to have a nonlinear relationship with the response variable, it could easily analyze the unbalanced data such as random missing data. [Conclusions] The analysis results of nonlinear mixed-effect models are reasonable and easily to be explained, and the nonlinear mixed-effect models provide a new analysis idea for categorical repeated measurement data.