Skip to content

Mismatch of gamma stat NP index if using or not using DropBin

TRExFitter Version and Commit Hash

TRExFitter version: v1.4.0 from StatAnalysis,0.6.3

Description

Since the index of gamma stat NP is defined after DropBin, this can cause an issue in a specific case:

The situation is we want to drop all the events within a signal mass window (i.e. SR blinding), but use the sideband data to get the best-fit NPs, and use them for Asimov data generation to get an expected sensitivity. You can find an example in this presentation. (plot on the right in P2)

We need to drop bins in the blinded SR data fit, while we don't drop when doing Asimov, as it needs to be generated in the full range with the injected signal. The way I handled it is to run 2 times trexfitter with and w/o using DropBin. This means

  • gamma_stat_bin_6 would correspond to the 12th bin if we drop those near the Higgs peak for the data-blinded fit
  • gamma_stat_bin_6 would correspond to the 7th bin at 125 GeV for the Asimov

Therefore, when I told the trexfitter to generate an Asimov, I had to manually tell it to use the fitted value gamma_stat_bin_6 for gamma_stat_bin_11, else we would see a pull of gamma around the signal.

Reproducer Configuration

(Let me know if a reproducer is really needed.)

Ideas to Resolve

I don't know if it will be possible to define the index of gamma_stat using the binning before DropBin.