SciPost Submission Page
Searches for new physics in collision events using a statistical technique for anomaly detection
by S. V. Chekanov
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | Sergei Chekanov |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2110.06277v1 (pdf) |
| Date accepted: | 2022-02-28 |
| Date submitted: | 2021-10-14 16:19 |
| Submitted by: | Chekanov, Sergei |
| Submitted to: | SciPost Physics Proceedings |
| Proceedings issue: | 50th International Symposium on Multiparticle Dynamics (ISMD2021) |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approaches: | Experimental, Phenomenological |
Abstract
This paper discusses a statistical anomaly-detection method for model-independent searches for new physics in collision events produced at the Large Hadron Collider (LHC). The method requires calculations of Z-scores for a large number of Lorenz-invariant variables to identify events that deviate from those expected for the Standard Model (SM).
Published as SciPost Phys. Proc. 10, 015 (2022)
Reports on this Submission
Report #1 by Anonymous (Referee 1) on 2022-1-28 (Invited Report)
- Cite as: Anonymous, Report on arXiv:2110.06277v1, delivered 2022-01-28, doi: 10.21468/SciPost.Report.4264
Strengths
A clearly written paper, giving specifics of the simple statistical method and an example of good performance.
Weaknesses
Appropriate for proceedings.
Report
While not a block for proceedings publication, I note:
0) The RMM is introduced in the author's previous papers, but not really explained here despite not being a widely known concept. But there was a page limit...
1) The effective assumption of Gaussian X_i distributions for RMM seems a bit dangerous, since non-Gaussian tails could masquerade as BSM outliers. I think this can be remedied by a more explicit calculation of full empirical p-value from the found distribution, and conversion to Z, though.
2) The Fig 1 performance plots seem equally normalised across samples with very different cross-sections... this isn't quite clear from the presentation, and a log plot with more realistic cross-sections would illustrate the physics power better, especially re. how the BSM distribution compares to the SM tails.
Requested changes
None needed.