ADA presentation Soheun Yi
Model-Agnostic Detection of New Physics using Data-Driven Signal Regions
Usually test for presence of signal with:
- known signal model
- known background model
Often dont have a precise background model
- need data-driven background (assumption)
EXAMPLE of sideband fit 1D Example in 2D
Here also want to relax the assumption that we know the signal !
Assume:
- signal is localized.
- more signal S/sqrt(B) in the 4b data
EXAMPLE toy 1D
- Cutting on p4/p3 is sensitive to background mismodeling
- Instead use low-pass filter
- SvB ratio of two FvT / smeared-FvT