多元logistic回归的共线性分析

Multivariate logistic Regression Collinearity Diagnosis Analysis.

目的 探讨多元logistic回归分析中共线性诊断的评价指标。方法 通过对logistic模型求解算法的分析 ,把线性回归中的共线诊断工具推广至非线性 ,详细介绍了评价过程、诊断标准及解决途径 ,最后应用到儿童奶瓶龋影响因素分析的实例中加以验证。结果 获得了logistic回归模型中关于共线性的重要信息。通过实例对详细过程进行了解答。结论 诊断方法能有效地检测到共线关系 ,为进一步采取备择方法提供准确依据 ,从而避免了错误结论的得出。

Objective In logistic model,collinear relationship between the independent variables can yield unstable and not easily interprered regression estimation.This paper explored the assessing indices of collinearity diagnosis in logistic regression.Methods Three tools for diagnosing collinearity in linear model were extented to nonlinear logistic model,by analysing Newton-Raphson iterative method which was used to solve logistic model.This extention is based on the information matrix of logistic model.Conclusion this technique is applied to the logistic regression model that related the children's milk bottle caries to its risk factors,the results show that these measures can effectively detect the collinear relationships and its magnitude.