摘要
目的：本研究试图解决多个异常点情况下经典回归诊断方法中存在的掩盖现象。方法：采用ＬＭＳ与ＭＶＥ估计后求得的残差与稳健的马氏距离作为诊断量，且均以重复抽样算法实现。结果：该法得到的诊断量明显优于传统方法，且可作出直观的诊断图示。结论：此法可有效地克服医学回归数据诊断中的掩盖现象，值得推广。
Objective:this research is to deal with the masking effect when multiple outliers occur during the traditional diagnostic regression analysis.Methods:using LMS and MVE estimators to obtain robust residual and a robust Mahalanobias distance as diagnostics,and both are got by a kind of resampling algorithm.Result:The dignostics from this method is apparently better than the traditional method,and it can present the result in a graphic way.Conclusion:this method can conquer the masking problems effectively in the regression diagnostic analysis during the medicine research,so it is worth applying forward.
出处
《中国卫生统计》
CSCD
北大核心
1998年第5期1-4,共4页
Chinese Journal of Health Statistics
基金
山西省归国人员基金
全国统计科学研究计划项目
关键词
回归诊断
多个异常点
稳健估计
线性回归
Regression diagnosis Multiple outliers Masking Robust estimation
