Holle R, Windeler J
GSF-Forschungszentrum für Umwelt und Gesundheit, Neuherberg, Germany.
J Clin Epidemiol. 1997 Jan;50(1):117-20. doi: 10.1016/s0895-4356(96)00311-3.
Several authors have proposed alternatives to sensitivity and specificity which they recommend as so-called "chance-corrected" versions of these parameters of diagnostic validity. We argue that these new measures have some undesirable properties in comparison with the established measures and no substantial advantages. In particular, the extension of this concept to chance-corrected ROC curves is shown to be less useful than classical ROC analysis.
几位作者提出了灵敏度和特异度的替代指标,并推荐将其作为诊断效度这些参数的所谓“机会校正”版本。我们认为,与既定指标相比,这些新指标存在一些不良特性,且没有实质性优势。特别是,事实证明,将这一概念扩展到机会校正的ROC曲线不如经典的ROC分析有用。