Skip to content

Fix 4fs/5fs overlap, support systematic samples

Sebastien Wertz requested to merge variousChanges into master
  • Fix the logic for 4FS/5FS overlap in plotIt. Somes comments though:
    • Currently there is no overlap because the "ttB" component of the 5FS samples is never produced (see samples_template.yml). However at some point we'll need to produce both (control plots, unfolding checks etc.).
    • Please double-check if it makes sense now ^^
    • This only affects what is propagated to the plots.yml config used by plotIt, not what jobs are submitted.
  • Support systematic variation samples: for now, hdamp, tune, CR1, CR2, erdON (5FS samples only)
    • hdamp variations for 4FS yet to come
    • For the 4FS sample we'll take the tune and CR relative uncertainties from the 5FS ones (TODO)
    • CR1, CR2, erdON are simple alternatives and don't have up/down variations, so they can't be used to plotIt. We also probably can't simply put them as nuisance parameters in the fit!
    • The base sample name in the sample template for a systematic SYST relative to the nominal sample NOM (with subprocess SUB and era ERA) is renamed to "NOM_SUB__ERA__SYST" when filling the sample template: this is required for plotIt to be able to use take the up/down histograms from separate files instead of from the nominal file.
    • There is an "issues" with the normalization: plotIt will normalize the histograms (xs*lumi/sumW) by taking the sum of weights from the nominal sample, but that will be different from the systematic sample. Here I've added a function in the postprocessing utils to "correct" the normalization of the systematic histograms, and to copy those into the output ROOT file of the nominal sample. The original systematic histograms are not affected. PlotIt will give precedence to the systematic histograms found in the nominal file to those in the systematics files.
Edited by Sebastien Wertz

Merge request reports