Best estimation of a binary outcome in the sense of KS and AROC

被引：1

作者：

Appel, MJ ^{[1
]}

机构：

[1] Mortgage Guaranty Insurance Corp, Milwaukee, WI 53201 USA

来源：

COMMUNICATIONS IN STATISTICS-THEORY AND METHODS | 2002年 / 31卷 / 08期

关键词：

AROC; ROC curve; Kolmogorov-Smirnov separation; regression;

D O I：

10.1081/STA-120006066

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

Let (Z, Y) be a real random vector with Y binary. It is shown that the conditional probability f*(Z) = Pr(Y = 1\sigma(Z)) maximizes uniform Kolmogorov-Smirnov separation of false from true negative rates over all Q(Z)-measurable functions. An inequality is derived which provides conditions under which f* also maximizes the area under the receiver operating characteristic (AROC).

引用

页码：1251 / 1257

页数：7