Low and 0 MC prediction warnings/messages #527
Description
Is your feature request related to a problem? Please describe.
Having analysis sample binnings allowing for very low to 0 MC predictions is potentially dangerous both in case of "nominal" model/predictions, or anywhere in the parameter space during the analysis sampling. If not for sure motivated by physics, 0 MC predictions are an issue for the standard implemented statistics tests (as dealt with in #520) and might cause inconsistencies.
Describe the solution you'd like
User should be notified about the occurrences of low or 0 MC predictions in the bins of nominal and also reweighted analysis sample PDFs to double check they are well motivated.
Describe alternatives you've considered
The only alternative we have discussed so far would be stopping or throwing from the analysis, unless reassured or fixed the empty bins.
Better to allow running the analysis, while informing on the issue.
Additional context
Taking into account computational needs of a concrete implementation proposal, these info messages shall either inform about the fact it happened sometime during the analysis (a counter maybe), or provide extra information if requested like bin information, or reweighting parameter values (too much?).