[Rivet] WFinder

[Chris Pollard]

Hi Jon,

Is the "total missing four-vector" counted before or after hadronic decays?
>
> If it's before (as I hope it is), then surely for the samples the
> experiments actually *used* in their unfolding, the "truth MET" and
> "neutrino ET" are identical, so both are ok and the measurements are fine.
> But if (in nature or a new MC process) a different particle gives rise to
> that MET, the WFinder gives an unphysical result, which can't be compared
> properly to data, whereas using truth MET would be robust against this.
>

Yes: non-BSM invisibles "break" the WFinder very badly in its current
implementation, and that's a problem. But there are also other issues here,
too. Somewhat contrived example: the fiducial MET definition may not
include the forward regions that can be important (e.g. VBS/F, t/u channel
processes). In this case the sum of the prompt neutrino four momenta may
not approximate the fiducial missing momentum very well.

Unfortunately many analyses have already unfolded to the "hardest prompt
neutrino" instead of the fiducial MET. I don't think there's any way to
undo this now, but if we have a WFinder it should do something more
sensible.

Given that, I'm puzzled what "signal efficiency dropping massively" means.
> I would hope the change makes almost no difference (though of course it
> would require re-validation). I think having a WFinder which relies on
> physics, pre-hadron-decay observables something rivet needs...
>

I also don't understand the drop in efficiency. David, do you have a handle
on that?

Chris
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.hepforge.org/lists-archive/rivet/attachments/20150706/838d6304/attachment.html>