A comparison of two parametric ROC curves estimators in binormal model

Investor logo

Warning

This publication doesn't include Faculty of Arts. It includes Faculty of Economics and Administration. Official publication website can be found on muni.cz.
Authors

MICHÁLEK Jaroslav DOUDOVÁ Lucie SEDLAČÍK Marek

Year of publication 2005
Type Article in Proceedings
Conference Proceedings of the 23rd International Conference Mathematical methods in Economics 2005
MU Faculty or unit

Faculty of Economics and Administration

Citation
Field Applied statistics, operation research
Keywords ROC curve; CDF; empirical CDF; the weighted regression estimator; simulations
Description The receiver operating characteristic curve ROC(t) plays an important role in many economic branches when it is necessary to classify between two populations. There are several approaches how to estimate it: classical estimator based on the empirical CDF, the weighted regression estimator, the estimator based on the best unbiased CDF estimator, non-parametric approaches based on the kernels or bootstraping. In the contribution the binormal model is considered and the attention is concentrated on the comparison of two parametric methods. The first one is the method of reweighted least squares where weights depend on the response variable. The second one is based on the best unbiased CDF estimator. The comparison of both methods will be demonstrated by simulations and the performance of both methods will be discussed for small sample sizes.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.