离散型二分类数据的广义混合线性模型分析——混合线性模型及其SAS软件实现(三)

Generalized Mixed Linear Model for Discrete Binary Data-Mixed Linear Model and its Implementations of SAS(3).

目的 离散型系统结构数据若忽略其误差多层次的存在 ,采用广义线性模型分析会导致不实的结果 ,本文探讨了广义混合线性模型理论及其实现过程。方法 利用SAS宏程序GLIMMIX实现了logit及loglog混合模型的分析。结果 各项结果均与MLN的分析结果一致 ,且进一步拟合了数据的logit随机系数模型。 结论 GLIMMIX可分析多种离散型系统结构数据 ,但还应注意算法和数据的误差分布 ,才能使广义混合线性模型得以合理地应用。

Objective Ignoring the hierarchical error of discrete hierarchical structured data,generalized linear model can give the incorrect results.The theory and implementation of generalized mixed linear model was discussed.Methods Perform the logit and loglog mixed model using SAS GLIMMIX macro.Results All the outputs accorded with those of MLN, in addition, random coefficient model was fitted with the data.Conclusion We should give attention to the algorithm and error distribution for data, so that generalized mixed linear model got the reasonable application.