consider 1 case study, pick a selection of 2-3 observables, perform a selection of MC variations
for a given setup X=(FO/matching/merging accuracy, PS model, NP model), produce envelope
check if the envelope from setup X behaves as expected or not: for instance, it might happen that even in a region that should be dominated by “hard physics”, one has a residual dependence upon the NP model, or a too-large dependence upon some shower parameter, if the variations are not done properly, or if there's an inconsistency.
check if envelopes from setup X_i and setup X_j overlap at least partially. If there are very large differences, is this expected (given the kinematic region probed and the perturbative/non-perturbative content of the setups)?
Disclaimer
to some extent, such a study could be considered a continuation of what can be found in chapter V.1 of https://arxiv.org/abs/1605.04692 . Nowadays we might aim for having higher accuracy in the ME part, and possibly more developed and more efficient frameworks for “PS-reweighting”.
Possible test cases:
pth in gluon-fusion (pros: directly related to STXS bins)
Drell-Yan (pros: there's data available)
ttbar (pros: atlas and cms have already several studies performed with identical or very similar setups)
Meeting on Monday
agree first on the utility/need (from the EXP side) of such a study, and then on the best test case
→Check
Decisions
Agreement reached points towards ttbar as first choice. Probably as a second choice: gg→H
Next steps
Work out a strategy (slack?)
Interested participants (please add your name):
* ttbar
* Emanuele Re
* Helen Brooks (can provide Vincia+POWHEG, maybe Pythia)