During the Osc workshop it came to our attention that the geant4reweight calculators use uncorrelated 20% uncertainties on all processes. @grayputnam reached out to authors who replied with this message:
The 20% isn't based on any data, it's just a default value as far as I know. The reasoning in microboone was that if the G4 uncertainties are very small relative to other sources with this 20% value (which they are), there's probably not much point in trying to precisely tune the inputs.
Inputs need tuning. sbncode implements the Geant4WeightCalc.cxx calculator here. It is based on a multisim approach.
Suggestion: dedicated study pulling in geant4reweight + sbncode + experiment code. Implement the correlation matrix from the g4rw paper, 2105.01744 (see attached image).
During the Osc workshop it came to our attention that the
geant4reweightcalculators use uncorrelated 20% uncertainties on all processes. @grayputnam reached out to authors who replied with this message:Inputs need tuning. sbncode implements the
Geant4WeightCalc.cxxcalculator here. It is based on a multisim approach.Suggestion: dedicated study pulling in
geant4reweight+ sbncode + experiment code. Implement the correlation matrix from the g4rw paper, 2105.01744 (see attached image).