广义线性混合模型及其SAS实现.pdfVIP

第10 期 统计教育 No. 10 （总第121 期） (Series No. 121) 2009 年10 月 Statistical Thinktank Oct 2009 广义线性混合模型及其SAS 实现 康萌萌 摘 要：本文探讨了重复测量资料广义线性混合模型(GLMMs)建模及SAS9.1 的PROC GLIMMIX 程序实现。 利用 PROC GLIMMIX 程序中Model 语句选项和Link 语句来指定因变量的分布及连接函数， 通过Random 语句来指定 随机效应，采用线性限制性 残差虚拟似然法进行参数估计。 是在广义线性模型的基础上引入随机效应，因 / GLMMs 变量可以是指数家族中任意分布，可以通过连接函数将观测的均数向量与模型参数联系起来。 GLMMs 应用范围 广，建模灵活，可以为相关或非常数方差数据建模，能提供客观正确的统计结论。 广义线性模型；广义线性混合模型； 关键词： SAS PROC GLIMMIX Generalized Linear Mixed Models and Implementation with SAS Kang Mengmeng Abstract: This paper discusses generalized linear mixed models (GLMMs) of repeated measurement data, implementation with PROC GLIMMIX in SAS9.1 software. Using the MODEL and LINK sentences of PROC GLIMMIX to indicate the distribution and link function of the response, adopting the RANDOM sentences to indicate random effects, the paper makes use of restricted pseudo -likelihood or residual pseudo -likelihood based on linear to estimate the model parameters. GLMMs are the extension of generalized linear modeling by including the random effects. The response may be one of the exponential families and the vector of expected means of the observations is linked to the model parameters by a link function. GLMMs have wide implicational fields, can easily fit statistical models to data with correlations or no constant variance, and provide the objectively correct statistical results. Key Words: generalized linear model, generalized linear mixed model, SAS PROC GLIMMIX 1972 年，Ne