Improved methods for bandwidth selection when estimating ROC curves.

DOI10.1016/S0167-7152(03)00150-0zbMath1113.62316WikidataQ58297520 ScholiaQ58297520MaRDI QIDQ1423137

Rob Hyndman, Hall, Peter

Publication date: 14 February 2004

Published in: Statistics \& Probability Letters (Search for Journal in Brave)

Full work available at URL: https://doi.org/10.1016/s0167-7152(03)00150-0

zbMATH Keywords

curve; Bandwidth selection; Kernel estimator; MISE; ROC; Binary classification

Mathematics Subject Classification ID

62G07: Density estimation

Related Items

Unnamed Item, A new method of kernel-smoothing estimation of the ROC curve, Optimal adaptive estimation of the relative density, Sequential estimate of partial area under ROC curve with \(\beta \)-protection and optimal ratio of cases to controls, Nonparametric estimation of the ROC curve based on smoothed empirical distribution functions, Constructing the best linear combination of diagnostic markers via sequential sampling, Compare diagnostic tests using transformation-invariant smoothed ROC curves, Nonparametric estimation of conditional ROC curves: application to discrimination tasks in computerized detection of early breast cancer, Modelling receiver operating characteristic curves using Gaussian mixtures, Pointwise exact bootstrap distributions of ROC curves, Minimum distance estimation of the binormal ROC curve, Smooth estimation of ROC curve in the presence of auxiliary information, Empirical process approach to some two-sample problems based on ranked set samples, Confidence bands for ROC curves, Non-parametric smoothed estimation of multivariate cumulative distribution and survival functions, and receiver operating characteristic curves, ROC Curves in Non-Parametric Location-Scale Regression Models, Relative density estimation for left truncated and right censored data

Cites Work

• Unnamed Item
• Local data-driven bandwidth choice for density estimation
• Bandwidth selection for kernel distribution function estimation
• Kernel estimators of the ROC curve are better than empirical.
• Smoothing parameter selection for smooth distribution functions
• ROC curve estimation based on local smoothing
• Bandwith selection for the smoothing of distribution functions