Accuracy and cut-off point selection in three-class classification problems using a generalization of the Youden index

Christos T. Nakas, Todd A. Alonzo, Constantin Yiannoutsos

Research output: Contribution to journalArticle

44 Citations (Scopus)

Abstract

We study properties of the index J3, defined as the accuracy, or the maximum correct classification, for a given three-class classification problem. Specifically, using J3 one can assess the discrimination between the three distributions and obtain an optimal pair of cut-off points c1<c2 in the sense that the sum of the correct classification proportions will be maximized. It also serves as the generalization of the Youden index in three-class problems. Parametric and non-parametric approaches for estimation and testing are considered and methods are applied to data from an MRS study on human immunodeficiency virus (HIV) patients.

Original languageEnglish
Pages (from-to)2946-2955
Number of pages10
JournalStatistics in Medicine
Volume29
Issue number28
DOIs
StatePublished - Dec 2010

Fingerprint

Classification Problems
Virus
Discrimination
Proportion
Testing
HIV
Class
Generalization
Human

Keywords

  • Accuracy
  • Cut-off selection
  • Maximally selected statistics
  • Random walks
  • ROC analysis
  • True-class rate
  • Youden index

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Accuracy and cut-off point selection in three-class classification problems using a generalization of the Youden index. / Nakas, Christos T.; Alonzo, Todd A.; Yiannoutsos, Constantin.

In: Statistics in Medicine, Vol. 29, No. 28, 12.2010, p. 2946-2955.

Research output: Contribution to journalArticle

@article{56446f3447b64ad3966377bfdc912cc5,
title = "Accuracy and cut-off point selection in three-class classification problems using a generalization of the Youden index",
abstract = "We study properties of the index J3, defined as the accuracy, or the maximum correct classification, for a given three-class classification problem. Specifically, using J3 one can assess the discrimination between the three distributions and obtain an optimal pair of cut-off points c12 in the sense that the sum of the correct classification proportions will be maximized. It also serves as the generalization of the Youden index in three-class problems. Parametric and non-parametric approaches for estimation and testing are considered and methods are applied to data from an MRS study on human immunodeficiency virus (HIV) patients.",
keywords = "Accuracy, Cut-off selection, Maximally selected statistics, Random walks, ROC analysis, True-class rate, Youden index",
author = "Nakas, {Christos T.} and Alonzo, {Todd A.} and Constantin Yiannoutsos",
year = "2010",
month = "12",
doi = "10.1002/sim.4044",
language = "English",
volume = "29",
pages = "2946--2955",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "28",

}

TY - JOUR

T1 - Accuracy and cut-off point selection in three-class classification problems using a generalization of the Youden index

AU - Nakas, Christos T.

AU - Alonzo, Todd A.

AU - Yiannoutsos, Constantin

PY - 2010/12

Y1 - 2010/12

N2 - We study properties of the index J3, defined as the accuracy, or the maximum correct classification, for a given three-class classification problem. Specifically, using J3 one can assess the discrimination between the three distributions and obtain an optimal pair of cut-off points c12 in the sense that the sum of the correct classification proportions will be maximized. It also serves as the generalization of the Youden index in three-class problems. Parametric and non-parametric approaches for estimation and testing are considered and methods are applied to data from an MRS study on human immunodeficiency virus (HIV) patients.

AB - We study properties of the index J3, defined as the accuracy, or the maximum correct classification, for a given three-class classification problem. Specifically, using J3 one can assess the discrimination between the three distributions and obtain an optimal pair of cut-off points c12 in the sense that the sum of the correct classification proportions will be maximized. It also serves as the generalization of the Youden index in three-class problems. Parametric and non-parametric approaches for estimation and testing are considered and methods are applied to data from an MRS study on human immunodeficiency virus (HIV) patients.

KW - Accuracy

KW - Cut-off selection

KW - Maximally selected statistics

KW - Random walks

KW - ROC analysis

KW - True-class rate

KW - Youden index

UR - http://www.scopus.com/inward/record.url?scp=78649575097&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78649575097&partnerID=8YFLogxK

U2 - 10.1002/sim.4044

DO - 10.1002/sim.4044

M3 - Article

VL - 29

SP - 2946

EP - 2955

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 28

ER -