广义线性混合效应模型在临床疗效评价中的应用

目的:探讨临床疗效评价中分类重复测量资料的广义线性混合效应模型(GLMMs)及SAS8.0的GLIMMIX宏实现。方法:利用GLIMMIX宏ERROR和LINK语句来指示疗效指标的分布及连接函数,通过REPEATED和RANDOM语句的TYPE选项选择合适方差-协方差结构矩阵来模拟不同时间疗效指标的相关性,采用基于线性的伪似然函数进行模型参数估计。结果:广义线性混合效应模型允许临床疗效评价指标是指数家族中任意分布(如:连续分布包括正态分布、beta分布、卡方分布等;离散分布包括二项分布、泊松分布、负二项分布等),可以通过连接函数将疗效指标的均数向量与模型参数建立线性关系,简化运算过程。结论:广义线性混合效应模型建模灵活,可为临床疗效评价提供更丰富的信息。

Apllications of Generalized Linear Mixed Models in Clinical Curative Effects Evaluation

Objective :To discuss generalized linear mixed models(GLMMs) of categorical repeated measurement datas in clinical curative effect evaluation,implementing with GLIMMIX macro in SAS8.0 soft.Methods: Using the ERROR and LINK sentences of GLIMMX macro to sign the distribution and link function of the index,adopting the TYPE option of REPEATED and RANDOM sentences to select the appropriate variance-covariance matrixs for modeling the relations,making use of pseudo-likelihood function based on linear to estimate the model parameters.Results: GLMMs allow the index may be one of the exponential family(Contimuum distributions including Nomal,beta distribution,chi-squared distribution etc;Disperse distributions including Binomal,Poisson and inverse Binomal etc),the vecor of expected means of the index is linked to the model parameters by a link function and model the linear equation, simple the calculator procedure.Conclusion: GLMMs can easily fit statistical models,the results are objective and reality,can strongly provide the abundant information for clinical curative effect evaluation.