IF


IF Document 144-v1

Neutrino Seminar Series: Unfolding Techniques: A Statistician’s Perspectiv

Document #:
IF-doc-144-v1
Document type:
seminar
Submitted by:
Evan Niner
Updated by:
Evan Niner
Document Created:
17 Apr 2017, 13:29
Contents Revised:
17 Apr 2017, 13:29
DB Info Revised:
17 Apr 2017, 13:29
Viewable by:
  • Public document
Modifiable by:
Abstract:
Any differential cross section measurement in experimental particle physics is smeared by the finite resolution of the particle detectors. Using the smeared observations to infer the true particle-level spectrum is an ill-posed inverse problem, typically referred to as unfolding or unsmearing. In this talk, I will first give an overview of the statistical techniques that are presently used to unfold particle spectra and give some practical recommendations based on experience from the LHC. I will then focus on the very challenging problem of quantifying the statistical uncertainty of the unfolded spectrum. I will demonstrate that techniques that are currently used may seriously underestimate the unfolded uncertainty. I will then propose two complementary approaches for obtaining improved unfolded confidence intervals. The first approach is based on debiasing the unfolded point estimate to obtain a bias-variance trade-off that is better suited for uncertainty quantification, while the second approach imposes shape constraints that are particularly relevant for unfolding steeply falling differential cross sections.
Files in Document:
Topics:
Authors:
DocDB Home ]  [ Search ] [ Last 20 Days ] [ List Users in Groups ] [ List Authors ] [ List Events ] [ List Topics ]

--------
Security, Privacy, Legal
Fermilab Policy on Computing
Fermi National Accelerator Laboratory

For assistance contact Service Desk



DocDB Version 8.7.23, contact Document Database Administrators