医学诊断试验评价的ROC分析——重复测量诊断数据的ROC曲线

The ROC Analysis for Evaluation of Diagnostic Accuacy in Repeated Measurements

目的探讨诊断试验中重复有序分类测量数据的ROC分析方法。方法在有序模型基础上构造相应的ROC模型 ,然后选用累积logistic模型作为联系函数 ,采用GEE方法进行参数估计。结果通过累积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。结论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。

Objective To explore the ROC analysis methods for repeated ordinal categorical data in diagnostic test. Methods ROC models are given based on ordinal regression models, and cumulative logistic models are used as the link function. The parameters can be estimated by using generalized estimating equations(GEE) method. Results ROC curve equation, areas under the ROC curve and standard error are estimated by the cumulative logistic models. Conclusion The methods in this paper can analyze and process effectively the repeated measures data in the image diagnostic test.