Controlling the False Discovery Rate via Competition: is the +1 needed?
Andrew Rajchert, Uri Keich
TL;DR
This paper investigates the necessity of the '+1' adjustment in the knockoff-based false discovery rate control method, questioning whether it is always required for accurate feature selection.
Contribution
The study critically examines the role of the '+1' in FDR estimation, providing insights into its necessity across different settings.
Findings
The '+1' may not be necessary in all scenarios.
Alternative methods can control FDR without '+1'.
Theoretical analysis clarifies conditions for '+1' requirement.
Abstract
Barber and Cand\`es (2015) control of the FDR in feature selection relies on estimating the FDR by the number of knockoff wins +1 divided by the number of original wins. We study the necessity of the +1 in general settings.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdversarial Robustness in Machine Learning · Explainable Artificial Intelligence (XAI) · Imbalanced Data Classification Techniques