对数线性模型的统计信息功能

Statistical Information Functions of Log-linear Model

通过一个四维列联表算例研究了对数线性模型在统计分析中的信息功能.结果表明:对数线性模型可通过拟合优度统计检验发现多维分类变量中存在的统计独立关系;可通过参数估计分别针对列联表的每一个组格回答各个单个变量以及2个、3个、4个变量的各种交互对形成该组格概率的贡献;无法提供变量间统计相依程度强弱的信息.

The paper discusses the information function in statistical analysis of log-linear models through an example of four dimension contingency table. Research shows that log-linear models can help find out the statistical independence relationship that exists in the multidimensional classification variables through a goodness-of-fit statistic test;log-linear models can help work out the lattice probability contribu tion of various interaction group formed by individual variable as well as two, three, four variables through the parameter estimation for each group lattice of contingency table log-linear models can not provide strength or weakness information of statistical dependence between variables.