A Note on the Area under the Gains Chart
DOI:
https://doi.org/10.6000/1929-6029.2018.07.03.1Keywords:
ROC curve, gains chart, 2AFC.Abstract
The Receiver Operating Characteristic (ROC) chart is well known in medicine and machine learning. In particular the area under the ROC chart measures the probability of correct selection in a two alternative forced choice (2AFC) scenario. The gains chart is closely related to the ROC curve but carries extra information about the rate at which the classifier identifies response, information that is not carried by the ROC chart. In this note, we point out that the appropriate area under the gains chart is identical to the analogous area under the ROC chart and that the gains chart is therefor to be preferred as a summary of classifier success.
References
Bamber D. The area above the ordinal dominance graph and the area below the receiver operating graph. Journal of Mathematical Psychology 1975; 12: 387. https://doi.org/10.1016/0022-2496(75)90001-2 DOI: https://doi.org/10.1016/0022-2496(75)90001-2
Dorfmann DD, Alf AE. Maximum likelihood estimation of parameters of signal detection theory - a direct solution. Psychometrika 1968; 33: 113. https://doi.org/10.1007/BF02289677 DOI: https://doi.org/10.1007/BF02289677
Hsieh F, Turnbull BW. Non-parametric and semi-parametric estimation of the receiver operating characteristic curve. Ann Statistics 1996; 24: 25. https://doi.org/10.1214/aos/1033066197 DOI: https://doi.org/10.1214/aos/1033066197
Lloyd CJ. The use of smoothed ROC curves to summarise and compare diagnostic systems. Journal of the American Statistical Association 1998; 93: 1356. https://doi.org/10.1080/01621459.1998.10473797 DOI: https://doi.org/10.1080/01621459.1998.10473797
Lloyd CJ, Zhou Y. Kernel estimators of the ROC curve are better than empirical. Statistics and Probability Letters 1999; 44: 221. https://doi.org/10.1016/S0167-7152(99)00012-7 DOI: https://doi.org/10.1016/S0167-7152(99)00012-7
Wieand S, Gail M, James B, James K. A family of non-parametric statistics for comparing diagnostic markers with paired and unpaired data. Biometrika 1989; 76: 585. https://doi.org/10.1093/biomet/76.3.585 DOI: https://doi.org/10.1093/biomet/76.3.585
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2018 Chris J. Lloyd
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Policy for Journals/Articles with Open Access
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are permitted and encouraged to post links to their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work
Policy for Journals / Manuscript with Paid Access
Authors who publish with this journal agree to the following terms:
- Publisher retain copyright .
- Authors are permitted and encouraged to post links to their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work .
