具有协变量或干扰因素的诊断试验数据的ROC分析  

The ROC Analysis for Diagnostic Test Data with Covariates

作　　者：李康[1] 马葆华[1] 赵亚双[1] 李振山[2] 

机构地区：[1]哈尔滨医科大学卫生统计教研室,150001 [2]哈尔滨理工大学电子工程系

出　　处：《中国卫生统计》2002年第2期67-70,共4页Chinese Journal of Health Statistics

基　　金：国家自然科学基金资助项目 (30 0 4 0 0 39)

摘　　要：目的　探讨具有协变量或干扰因素的诊断试验数据的评价问题 ,建立相应的统计模型及分析方法。方法　基于ROC分析给出多变量ROC模型 ,采用有序logistic连接函数 ,利用SAS软件进行参数估计 ,得到有协变量情况下的ROC曲线方程及曲线下面积。文中结合检测动脉硬化的数据 ,说明了其用法。结果　采用文中给出的方法 ,可以有效地扣除协变量的影响 ,准确地评价诊断的作用。结论　本文提供的方法能够有效地解决诊断试验中具有干扰因素影响的问题。Objective To explore the assessment techniques for data of diagnosis test with covariates and interference factors, to establish the statistics models and analyzing methods correspondingly.Methods The multiple variable ROC models were given based on ROC analysis, using ordinal logistic link function and SAS software to estimate parameters. The ROC curve and corresponding areas under ROC curve with covariates were obtained. Our method is illustrated by application to detect atherosclerotic data.Results Using the method can get rid of the influence of covariates effectively, thus evaluating the performance of diagnostic accurately. Conclusion The method provided in this paper can effectively resolve the problems of the influence of interference factors in the diagnostic test.

关 键 词：诊断试验 ROC分析 有序logistic模型 协变量 

分 类 号：R195[医药卫生—卫生统计学]