Cox模型影响点识别诊断统计量的模拟研究

目的 探讨Cox比例风险模型影响点的有效识别方法 .方法 通过Monte-Carlo模拟研究,比较了六种诊断统计量及相应诊断图对影响点的诊断效能.结果 模拟研究表明,加权score残差、似然距离和最大影响曲率及其诊断图可从模型拟合不同角度有效识别影响点.结论 影响分析应是Cox模型分析的一个重要组成部分,加权score残差、似然距离和最大影响曲率是有效识别影响点的诊断统计量.

Simulation Study of Diagnostics to Identify Influential Cases in Cox''''s Model

Objective To explore effective methods to identify influential cases in Cox's proportional hazards regression model.Methods By Monte-Carlo simulation,the power of six diagnostics to identify influential cases was presented.Results The simulation showed that weighted score residuals,likelihood displacement,maximum influence curvature and their plots can detect influential cases in Cox's model from different points of view.Conclusion Influence analysis should be an important part of the Cox's model.Weighted score residuals,likelihood displacement and maximum influence curvature are effective diagnostics to identify influential cases.