医学诊断试验评价的ROC分析_重复测量诊断数据的ROC曲线_李康.pdfVIP

2001 3 8 1 ROC * ROC 李 康 魏 韦 王滨友 赵亚双 探讨诊断试验中重复有序分类测量数据的ROC 分析方法 在有序模型基础上构造相 应的ROC 模型, 后选用累积logistic 模型作为联系函数, 采用GEE 方法进行参数估计 通过累积logistic 模型的参数估计, 能够得到ROC 曲线方程ROC 曲线下面积及标准误 上述方法能够对影像诊断试验中常 见的重复测量数据进行有效地分析与处理 诊断试验 ROC 曲线 重复测量 logistic 模型 The ROC Analysis for Evaluation of Diagnostic Accuacy in Repeated Measurements Li K ang, Wei Wei, Wang Biny ou , et al, H arbin Medical University, H arbin 150001 Abstract Objective To explore the ROC analysis methods for repeated ordinal categorical data in diagnostic test. Methods ROC models are given ased on ordinal regression models, and cumulative logistic models are used as the link funct ion. The parameters can e estimated y using generalized estimating equations( GEE) method. Results ROC curve equat ion, areasunder the ROC curve and standard error are estimated y the cumulative logist ic models. Conclusion T he methods in this paper can analyze and process effectively the repeated measures data in the image diagnost ic test. Key words Diagnostic test ROC curve Repeated measurement Logistic models , D , - 1 1, D = - 1 , , , , , D= 1 ; , Yk ( k = 1, 2, , m ) g ROC , , , P ( Y c | D , X ) = F ( + D + X + XD ) c , ( 1) c = 1, 2, , g - 1, ; c , , , ROC , ; ROC X = ( X , X , , X )p , = 1 2 p , logistic ( 1, 2, , p ) p T P( true- neg