scikit-hep
Binned chi2 zero bins
If I am fitting a Gaussian distributed variable with BinnedChi2, I am seeing a ValueError if one of the bins in the tails happens to be zero. The method throws this ValueError when the bin error is less than a small threshold value, but does not check to see if the bin weight is also zero. Can this case be checked before bringing the whole thing crashing down?
@WMGoBuffs - Thanks for filing the issue.
probfit hasn't been very active, so don't expect someone to pick up the suggestion and implement the change, you'll probably have to do it yourself.
The first step would be to post a minimal example (ideally just a few bins) showing:
- the ValueError you get now.
- mentioning the result you'd like to obtain for that case.
Once it's clear what exactly should happen, we can discuss if it's sensible, and then probably implementing it isn't very hard.
