Comparison of bandwidth selection methods for kernel smoothing of ROC curves

Xiao Hua Zhou, Jaroslaw Harezlak

Research output: Contribution to journalArticle

35 Citations (Scopus)

Abstract

In this paper we compared four non-parametric kernel smoothing methods for estimating an ROC curve based on a continuous-scale test. All four methods produced a smooth ROC curve of the test. The difference in these four methods lay with the way they chose their bandwidth parameters. To assess the relative performance of the four bandwidth selection methods, we conducted a simulation study using different underlying distributions, along with varied sample sizes. The results from our simulation study suggested that the kernel smoothing method originally proposed by Altman and Léger for estimation of the distribution function was the best choice for estimation of an ROC curve. We illustrated these methods with a real example.

Original languageEnglish
Pages (from-to)2045-2055
Number of pages11
JournalStatistics in Medicine
Volume21
Issue number14
DOIs
StatePublished - Jul 30 2002

Fingerprint

Kernel Smoothing
Bandwidth Selection
Receiver Operating Characteristic Curve
ROC Curve
Smoothing Methods
Kernel Methods
Simulation Study
Nonparametric Smoothing
Sample Size
Distribution Function
Choose
Bandwidth

Keywords

  • Bandwidth selection
  • Continuous-scale test
  • Empirical ROC curve
  • Kernel smoothing

ASJC Scopus subject areas

  • Epidemiology

Cite this

Comparison of bandwidth selection methods for kernel smoothing of ROC curves. / Zhou, Xiao Hua; Harezlak, Jaroslaw.

In: Statistics in Medicine, Vol. 21, No. 14, 30.07.2002, p. 2045-2055.

Research output: Contribution to journalArticle

@article{3a5397e196bb432aa22dc2baf8bcc46e,
title = "Comparison of bandwidth selection methods for kernel smoothing of ROC curves",
abstract = "In this paper we compared four non-parametric kernel smoothing methods for estimating an ROC curve based on a continuous-scale test. All four methods produced a smooth ROC curve of the test. The difference in these four methods lay with the way they chose their bandwidth parameters. To assess the relative performance of the four bandwidth selection methods, we conducted a simulation study using different underlying distributions, along with varied sample sizes. The results from our simulation study suggested that the kernel smoothing method originally proposed by Altman and L{\'e}ger for estimation of the distribution function was the best choice for estimation of an ROC curve. We illustrated these methods with a real example.",
keywords = "Bandwidth selection, Continuous-scale test, Empirical ROC curve, Kernel smoothing",
author = "Zhou, {Xiao Hua} and Jaroslaw Harezlak",
year = "2002",
month = "7",
day = "30",
doi = "10.1002/sim.1156",
language = "English",
volume = "21",
pages = "2045--2055",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "14",

}

TY - JOUR

T1 - Comparison of bandwidth selection methods for kernel smoothing of ROC curves

AU - Zhou, Xiao Hua

AU - Harezlak, Jaroslaw

PY - 2002/7/30

Y1 - 2002/7/30

N2 - In this paper we compared four non-parametric kernel smoothing methods for estimating an ROC curve based on a continuous-scale test. All four methods produced a smooth ROC curve of the test. The difference in these four methods lay with the way they chose their bandwidth parameters. To assess the relative performance of the four bandwidth selection methods, we conducted a simulation study using different underlying distributions, along with varied sample sizes. The results from our simulation study suggested that the kernel smoothing method originally proposed by Altman and Léger for estimation of the distribution function was the best choice for estimation of an ROC curve. We illustrated these methods with a real example.

AB - In this paper we compared four non-parametric kernel smoothing methods for estimating an ROC curve based on a continuous-scale test. All four methods produced a smooth ROC curve of the test. The difference in these four methods lay with the way they chose their bandwidth parameters. To assess the relative performance of the four bandwidth selection methods, we conducted a simulation study using different underlying distributions, along with varied sample sizes. The results from our simulation study suggested that the kernel smoothing method originally proposed by Altman and Léger for estimation of the distribution function was the best choice for estimation of an ROC curve. We illustrated these methods with a real example.

KW - Bandwidth selection

KW - Continuous-scale test

KW - Empirical ROC curve

KW - Kernel smoothing

UR - http://www.scopus.com/inward/record.url?scp=0037199385&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0037199385&partnerID=8YFLogxK

U2 - 10.1002/sim.1156

DO - 10.1002/sim.1156

M3 - Article

C2 - 12111886

AN - SCOPUS:0037199385

VL - 21

SP - 2045

EP - 2055

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 14

ER -