Statistics and Its Interface

Volume 10 (2017)

Number 2

Variable selection in ROC curve analysis with focused information criteria

Pages: 229 – 238



Baoying Yang (Department of Statistics, College of Mathematics, Southwest Jiaotong University, Chengdu, Sichuan, China)

Xin Huang (Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington, U.S.A.)

Gengsheng Qin (Department of Mathematics and Statistics, Georgia State University, Atlanta, Ga., U.S.A.)


In Receiver Operating Characteristic (ROC) curve analysis, many factors such as the study subject’s characteristics or operating conditions of a medical test may affect the diagnostic accuracy of the test. ROC regression models are introduced to accommodate effects of the covariates. If many covariates are available, variable selection problem arises. The area under the ROC curve (AUC) is a popular one-number summary index of the discriminatory accuracy of a medical test. In this paper, we propose a variable selection method based on the Focused Information Criteria (FIC) with focus on the AUC index. In particular, the FIC is developed in a placement-value model for ROC regression. The proposed method is illustrated through simulation studies and a real data example.


AUC, diagnostic test, placement value model, variable selection, ROC

Published 31 October 2016