/pyhf_tutorial

Primary LanguageJupyter Notebook

pyhf_tutorial

each file exists as a python script and a jupyter notebook

  • intro.py
    • generate toy example
    • create model dict
    • look at model some specs
    • do simple fit
    • draw profile likelihood
    • compare 3 different background normalisations (fixed, constrained, free)
  • fit.py
    • nuisance parameters and auxdata
    • generate toys
    • pull distributions
    • significance and upper limits
  • systematics.py
    • luminosity modifier
    • correlated shape modifier
    • uncorrelated shape modifier
    • toy study for model with wrong correlation assumption
    • model partially correlated uncertainties (eigendecomposition)
    • splitting uncertainties by systematic source
  • ratio.py
    • create toy MC and data for a second channel
    • fit independent signal strength in channel 2, share background normalisation parameter
    • rescale signal strength to correspond to BFs
    • compute ratio of correlated BFs
    • extract ratio directly as POI from the fit