Lee W C
Graduate Institute of Epidemiology, College of Public Health, National Taiwan University, Taipei, R.O.C.
Stat Med. 1999 Feb 28;18(4):455-71. doi: 10.1002/(sici)1097-0258(19990228)18:4<455::aid-sim44>3.0.co;2-a.
Several indices based on the receiver operating characteristic curve (ROC curve) have previously been found to possess probabilistic interpretations. However, these interpretations are based on some unrealistic diagnostic scenarios. In this paper, the author presents a new approach using the Lorenz curve. The author found that the summary indices of the Lorenz curve, that is, the Pietra index and the Gini index, can be interpreted in several ways ('average change in post-test probability', 'per cent maximum prognostic information', and 'probability of correct diagnosis'). These interpretations have a close tie with real-world medical diagnosis, suggesting that these indices are proper measures of test characteristics.
先前已发现基于接受者操作特征曲线（ROC曲线）的几个指标具有概率解释。然而，这些解释是基于一些不切实际的诊断场景。在本文中，作者提出了一种使用洛伦兹曲线的新方法。作者发现，洛伦兹曲线的汇总指标，即彼得拉指数和基尼指数，可以有几种解释（“检验后概率的平均变化”、“最大预后信息百分比”和“正确诊断的概率”）。这些解释与现实世界的医学诊断密切相关，表明这些指标是检验特征的合适度量。
