 |
 |
Thanks Vera for looking at this! > Thank you Kerstin for posting all these nice plots. > > My observation is the following: > - the dominant B to XC l nu background is not well described, > thus your background subtracted distributions look poor > in many cases. Yes. In some cases it is hard to say whether it is the background (or more specifically the background that passes the charm vetos) or the signal that is not well described, but I do agree that it is likely to be (at least also) the background in many cases. > - Of course the distributions that are the most difficult to > reproduce are multiplicities, both charged and neutrals, > in the recoil only, and also the total charge in the event > (which includes the Breco charge). > What are the cuts on the momentum/energy of charged > and neutral particles in the recoil? We have a table in the next version of BAD 1212 that lists the selection criteria. You can see it on page 6 of this unofficial version: http://costard.lbl.gov/~kerstin/vubunf/ana32/BAD1212_v5a.ps It is the selection proposed in BAD 1111 (but we picked it up from the BRecoilUser package, so we should not be affected by the typos in earlier version of that BAD). > The total charge distribution for the enriched sample > of neutral B gets very asymmetric in data. Is this a loss > of negative slow pions in data due to interactions? Is there a reason why we would expect to lose negative slow pions with higher probability than positive slow pions? And if so, would we think that for the charged B, the asymmetry is from slow pions that come from fragmentation processes or such (not from D* and so they would be in both B charge samples)? Spontaneously, I don't have an idea how we could test this easily. Shouldn't this be something we should know from the B0->D*lnu analyses? > the effect is much smaller in the charged B, for which > you do not apply the D* l nu veto. > - have you tried to change the purity of the Breco selection, > to see whether this might cause problems? We have changed purity cuts earlier in the analysis, but not recently, so we have not tried looking at this in this context. We could try this. > or enlarge the q2 cut on the recoil? Do you mean q2 = the invariant mass of the W? We do not apply any cut on q2. So are you suggesting we should try cutting on it? > - Do you understand the large change in charged multiplicity between > enriched and depleted samples? The neutral multiplicity increases > for the enriched samples, the charged multiplicity decreases. > Does this point to problems with the enriched MC? Jet-set > fragmentation? Are you looking at the background subtracted spectra or the spectra before background subtraction? I am not quite sure, but I think the trend is similar whether or not we look before or after background subtraction. I guess in the neutrals it could be JetSet fragmentation. Backgrounds from the detector should also show up in the depleted sample. Not sure how we could test this easily. > - the enriched Emiss-Pmiss and MM*2 distributions do not > show the expected peaks. Emiss-Pmiss= MM2/(Emiss+Pmiss) assumes > a linear dependence of the MM2 resolution on Emiss. For Emiss-pmiss I think it is fair to look at this after the mm2 cut (as we tend to think of the mm2 cut as cut to suppress background and Emiss-pmiss to then improve the mX resolution). The enriched sample then shows a good peak, the depleted is not so great. > Have you checked MM2 vs Emiss? Not on release 18. But we have studies we performed on release 14, which you can find here: http://costard.lbl.gov/~kerstin/vubunf/resolution/plots.html It does not have mm2 vs Emiss, though, but a bunch of other plots related to the neutrino (and also other things, resolution and truth matching studies). We could check mm2 vs. Emiss on the current ntuples. > I do not understand the motivation of the additional > cut on the MM^2. It's really more the Emiss-pmiss cut which is additional. The mm2 cut removes a lot of charm background at low mX (the long positive shoulder in the enriched sample). We showed this at the June CM: http://www.slac.stanford.edu/BFROOT/www/Organization/CollabMtgs/2007/detJun07/Tues2d/Tues2d.html Basically, mm2 suppresses charm background, Emiss-pmiss improves mX resolution. Kinematically it makes sense when you consider what losing a kaon does to these distributions. > - Have you tried to look at these distribution with fewer bins > on Mx? I am not sure I understand the question. Do you mean whether we have performed data-MC comparisons for a restricted range of mX? We have not tried that, since we are using the full range for the unfolding. But maybe we could learn from making the comparisons in the low mX only, as we would change the S/B significantly. Is that what you are thinking of? Kerstin > So many questions, no clear answers? > Ciao, > Vera > -----Original Message----- > From: [log in to unmask] [mailto:[log in to unmask]] On Behalf Of Kerstin Tackmann > Sent: Friday, August 31, 2007 2:26 PM > To: vub-recoil > Subject: Data-MC Agreement (again) > > > Dear vub-recoilers, > > as you probably know, we are seeing some problems in the data-MC > agreement. The most recent plots are here: > http://costard.lbl.gov/~kerstin/vubunf/ana32/dataMC/dataMC_KLcorr.html > > We obtain these plots as follows: > *Remove the cut on the given variable from the analysis. > *Perform the fit to the mX spectrum with this selection. > *Rebin the data and MC in the variable under study and use > the *comp from the previous fit to mX to scale the MC > components. > > The main concern is about the neutrino variables and Qtot. > > If we could have plots using this technique for the Vub analysis, that > would be a very interesting comparison, from which we could hopefully > learn and which would help us a lot in the review. I would really > appreciate if someone could look into this (as soon as possible) > -- please let me know if that would be possible. I'd be happy to give > more technical details, just let me know if something isn't quite clear! > > Thanks so much, > Kerstin > > >