有序结果变量回归及其应用

目的：介绍有序结果变量回归（Ordinal Regression）的基本原理,SPSS10．0的Ordinal Regression过程和确定“最佳”模型的方法。方法：介绍有序结果变量回归模型结构和模型评估指标,用一个实例和SPSS10．0的Ordinal Regression过程,说明有序结果变量回归的操作和结果分析。结果：列出五种连接函数两种模型的各种评估指标和总正确率,“最佳”有序结果变量回归模型的系数表。结果：有序结果变量回归适用于两个或以上分类水平的应变量和一组协变量。通过五种连接函数两种模型的各种评估指标和总正确率的比较,可挑选出“最佳”有序结果变量回归模型,并能预测某个体的分类。

Ordinal Regression and Its Application

Objective The paper introduces basic idea of ordinal regression, the ordinal regression procedure of SPSS10.0 and the method of deciding the "best" model.Methods The structure and evaluation of ordinal regression model are introduced. The operation and result analysis are illustrated with an example and the ordinal regression procedure of SPSS10.0.Results Each evaluation index and total correct rate of two models of five link functions and the table of coefficients of the "best" model are shown.Conclusion Ordinal regression allows you to model the dependence of polytomous ordinal response on a set of covariates. The "best" model can be selected through the comparison among each evaluation index and total correct rate of two models of five link functions and what group a case is can be predicted.