论文部分内容阅读
多维列联表资料是用来表达多个分类变量的计数资料,这些列联表资料可用对数线性模型进行拟合,用相似于回归分析中的逐步回归法找到一个最佳模型。由此得到对于观测频数起显著作用的主效应及其交互作用效应。本文主要介绍用后退法选择模型,据文献报道(Benedetti and Brown 1978)后退法是选择模型较好的一种方法。一、对数线性模型所谓模型是指关于观测结果的理论。在多元回归分析中最常用的是线性模型,应用最小二乘法或极大似然估计法确定模型中的
Multidimensional contingency table data is used to express the count data of multiple categorical variables. These contingency table data can be fitted by a logarithmic linear model and a best-fit model is found by stepwise regression similar to regression analysis. The main effects and their interaction effects that have a significant effect on the frequency of observations are thus obtained. This article mainly introduces the use of the back-up method to select the model. It is reported in the literature (Benedetti and Brown 1978) that the back-off method is a better method to select the model. First, the logarithmic linear model The so-called model refers to the observation of the theory. In the multiple regression analysis is the most commonly used linear model, the application of least squares or maximum likelihood estimation method to determine the model