 |
 |
Hi Francesca, -- I'll put this one to the vub-recoil list, too, so others can contribute if they wish... -- > > > as I mentioned at today's meeting, I just want to bring up how we did the > > > truth-match in the b->sgamma analysis (that we call indeed pseudo > > > truth-matching). > > > > > > The details are on BAD 768, v16, section 7.6. We distinguish between the > > > fully truth matched events (for which all the daughters are truth matched > > > to the B), and the onces which are not fully truth > > > matched (where the reconstructed track have an associated GHit, and the > > I am not sure I understand what you mean (I also read the part in BAD > > 786). I assume you mean you do not require every track to be truth > > matched using the standard truth matching (I think there a track is called > > truth matched if the true track caused the majority of the hits that > > are making up the reco track.). I guess you do *not* mean that your first > > category leaves out tracks once they have a single hit that is not > > truth matched to the right particle. Is this correct? > > The "have an associated GHit" is the formulation that is confusing me... > > yes, you are right. We say in the BAD 'associated GHit', but we actually > mean truthmatched track, with whatever is the number of GHits required. > > > > > > reconstructed final state is the same than the true final state), and for > > > which the reconstructed hadronic mass is within 50(25) MeV away from the > > > true one. > > > The sum of all the events makes the whole of the pseudo truth matched > > > events. > > > > > > In our case, this method worked well. Could it be feasible for the > > > semileptonic events? Do you see any drawback? > > > > Requiring the modes of the true and the reco'd B was actually our > > previous truth matching and we found this too strict for the > > 2 PDF mES fit, we need to be looser than this. > > > > What actually we had in b2sgamma are: > -fully truth-matched events: the tracks are truthmatched and asked to come > from the true B > - "pseudo-truthmatched" events: the tracks are truthmatched, the recoed > final states is asked to be identical to the true B final state (we do > not ask that the tracks come from the true B), a cut on > mX is applied. The events which fail the cut are our cross-feed > background, which mainly means a peaking background. > > We take all the events which pass at least one of the two criteria. > > We had at most 4 particles in the final state plus the gamma. > > > Concerning looking at the reco'd mass: > > So in our case we would look at the difference of the generated and > > reco'd mES if we wanted to follow this method as closely as possible? > > I guess we could try that, however, I wonder if it is a good idea. > > Loosely, this would mean to restrict the truth matching to about the > > mES signal region. While this would probably allow us to get a truth > > matching that looks nicer (I did look at this in December), we loose > > the mES tail for telling us how well we really do. > > > > Does this seem to make sense? > > I think one needs to look at the difference between the true and the > reconstructed mass and see how close there are. It could be that one > doesn't really lose the tails of the events if the cut is wide enough. BTW, the > cut should be dependent on the number of particles of the hadronic system. > > However, I really have the suspicion that if one applies this method will > necessarily have a peaking background, so most likely will need to go back > to the 3D PDF. Yes, probably. However, from the studies Antonio showed earlier this year we concluded that even on the high statistics samples the 3 PDF fit was unstable and if I am not mistaken we decided that we were going to abandon this approach... I think it makes more sense if we first fix our fit strategy and then find the best possible truth matching and not vice versa... > Kerstin: where can I find your study regarding the mass cut? Unfortunately, there is nothing I can point to. I just played with this a little and did not keep the results... sorry. While it looked nicer, it did not seem very sensible, since by just cutting at mES I was free to do whatever I wanted and it seemed too arbitrary to me. Your cut on the difference makes more sense to me, so we should rather try that I think. > > But I do think that an improvement of the current truth matching > > is more than welcome! > > > > Maybe it would be useful to move this discussion to the vub-recoil > > list to allow other people to contribute/follow. > > OK, I will forward all the emails to the mailing list. > > Unfortunately I will have another meeting starting at 8:00 on Tuesday so I > am not sure I will manage to attend the vub-recoil meeting. > If not possible, can I just call you and/or whoelse is interested > to have a chat on Tuesday after the meeting? I guess I would be > free only around 10:00-10:30. I have time until 11:00am. So we could do it until 11:00am, if we think that's enough time to start or we can do it on Wednesday or another day this week. Cheers, Kerstin