Closed finnmayhew closed 1 month ago
Looks good to me
The original reason for this treatment was a bias in the nutau appearance analysis when this error was not fixed. So maybe instead we could explicitly write here that this was needed to be fixed for Dragon nutau
The original reason for this treatment was a bias in the nutau appearance analysis when this error was not fixed. So maybe instead we could explicitly write here that this was needed to be fixed for Dragon nutau
This makes sense to me, will amend and do a new pull request.
Previously said: "Needed to allow mod_chi2 to behave correctly". My understanding is that this stage reduces the accuracy of mod_chi2 by approximating MC errors throughout the space as the initial MC errors.