SPSS 中的 ROC 分析用于检验/诊断方法的评价

目的]介绍如何用SPSS软件包中新颖的ROC分析法，来对一种或几种检验／诊断方法进行科学评价。方法]回顾目前文献中的存在问题；对四格表排列、数据库录入以及ROC命令默认设置提出应用或改变建议；举例说明定性、定量资料的RoC分析方法；介绍2或3种诊断方法诊断效能比较的RoC分析方法与结果；RoC分析结果与列联表分析、logistic比回归及判别分析结果相印证。结果]按本文建议，ROC分所能方便地对一种或几种检验／诊断方法进行科学评价，算出其敏感性与特异性等6项指标，且与其他3种分析方法统一、对应。结论]ROC曲线能将诊断方法的敏感性与特异性结合起来进行分析，而不是仅侧重于其敏感性或特异性，又能表示为“曲线下的面积越大，其诊断试验效果越好”，既全面又直观，又与其他统计分析方法结果对应、统一，值得在检验／诊断方法科学评价中广泛应用。

Application of ROC Analysis in SPSS for the Evaluation of Test / Diagnosis Methods

Objective\] To introduce the application of ROC analysis method in SPSS for evaluation of one or more test / diagnosis methods \ The problems were reviewed in recent articles;The arrange method of fourfold table, data input and the changing of default setting of ROC command were suggested Taking examples to illustrate the application method of ROC analysis, including qualitative and quantitative data;The comparison of ROC diagnostic effects of one or more diagnosis methods was introduced;The results of ROC analysis were compared and tested with other 3 methods (CROSSTABS, logistic regression and discriminant analysis in SPSS) \ According to the suggestions in this paper, the effect evaluations of one or more diagnostic methods were easily made by ROC analysis and the results were the same as the other 3 methods The 6 indicators (sensitivity and specificity and so on) of the method can be easily calculated \Conclusion\] The sensitivity and specificity of one diagnosis method can be combined by ROC analysis The area under the curve is shown clearly and the more, the better The results are the same as CROSSTABS, logistic regression and discriminant analysis in SPSS