Skip to content

Blinding issue in RooStats *allRegions* histograms

TRExFitter Version and Commit Hash

TRExFitter version: v4.26-dev Commit Hash: b5c0b4e8

Description

In the ttH(H->bb) legacy analysis, we define our STXS regions twice, once as a signal and once as a validation region (example given in Reproducer Configuration). The full config can be seen here.

I ran the n, w and d step, which produces the RooStats folder and everything correctly. In the RooStats folder, there are now two types of root files, Fit_1l_{REGION}_model.root and Fit_1l_allRegions_{REGION}_model.root files. While for the Fit_1l_{REGION}_model.root files, the blinding is correctly applied, the Fit_1l_allRegions_{REGION}_model.root files, which are explicitly declared as signal regions, the blinding is not applied.

The overall fit afterwards doesn't seem to be bothered by that, but the Yields.txt and the Table_prefit.yaml in the tables/ folder seem to be effected by this. Both regions, the signal and validation STXS regions, are unblinded.

This is also the case for the pre- and postfit summary plots. While the plots of the individual region are correctly blinded, the summary plots are not.

Reproducer Configuration 

Region: "Full_ttH_STXS1_1l"
  Type: VALIDATION
  Variable: "L2_Class_ttH_fraction/(L2_Class_tt1b_fraction+L2_Class_tt1B_fraction+L2_Class_tt2b_fraction)",3,0.0,0.6
  VariableTitle: "Ratio ttH/sum ttb comp. frac"
  Label: "STXS 1"
  NtuplePathSuff: "5j3b_ttH"
  # Binning: "AutoBin","TransfoD",2,2
  Binning: 0.000000,0.088620,0.116820,0.163440,0.600000
  RatioYmax: 1.4
  RatioYmin: 0.6
  Selection: L2_Reco_higgs_pt < 60000 && passedOfflineBoostedSelection == 0

Region: "ttH_STXS1_1l"
  Type: SIGNAL
  Variable: "L2_Class_ttH_fraction/(L2_Class_tt1b_fraction+L2_Class_tt1B_fraction+L2_Class_tt2b_fraction)",3,0.0,0.6
  VariableTitle: "Ratio ttH/sum ttb comp. frac"
  Label: "STXS 1"
  NtuplePathSuff: "5j3b_ttH"
  # Binning: "AutoBin","TransfoD",2,2
  Binning: 0.000000,0.088620,0.116820,0.163440,0.600000
  RatioYmax: 1.4
  RatioYmin: 0.6
  Selection: L2_Reco_higgs_pt < 60000 && passedOfflineBoostedSelection == 0
  AutomaticDropBins: TRUE

Ideas to Resolve

I don't really have an idea 😅

Tagging @alheld, @leevans

Edited by Alexander Froch