 |
 |
---------- Forwarded message ---------- Date: Thu, 04 May 2006 18:59:44 +0200 From: Concezio Bozzi <[log in to unmask]> To: Kerstin Tackmann <[log in to unmask]> Subject: Re: Question concerning correctionratiovxb Hi Kerstin, it seems I am drowning in a lot of emails and other things to do recently... I am sorry for that. Your remark on the B+/B0 correction ratio is correct, in the sense that it should be computed ideally without using any information on the recoil. In reality, this correction ratio is determined by using mes fits. If we don't ask for a lepton in the recoil we end up with a lot of background in the mes distribution, and the uncertatinty on the signal yields is dominated by the systematics of the mes fits. For this reason, we ask for the SL events selection, and this correction ratio is independent (at the first order) on which kind of events is in the recoil. Do you agree? Using the additional requirement on the flavor for neutral Bs is an unnecessary complication. It does not change the picture at first order since our correction ratio is a double ratio. The only thing one should make sure is that the composition of the recoil is as much as possible the same in MC and data. The problem I see in our code is that we have (Daniele says for historical reasons) two separate correction ratios for vub and vcb, whereas the correction ratio is universal, and in any case very close to the one we compute for vcb, with or without the flavor requirement. The reason for this is that the semileptonic events selection gives us a sample consisting mainly of vcb events in both data and MC. Do you agree? For this reason, the vub events should be indeed reweighted with the same ratio we use for vcb, and computing another correction ratio is plainly wrong. It turns out that the two correction ratios we compute are quite similar, and I regard this as an accident, since the composition of same and opposite flavor events in vcb and vub is as you say quite different. Are we making any progress? In any case, I must thank you for pointing out this. When I picked this analysis I took many things for granted and it's good that we check things again. Ciao, Concezio. On Wed, 2006-05-03 at 09:30 -0700, Kerstin Tackmann wrote: > Hi Concezio, > > sorry to bug you with this - I assume this just got lost... > I see that on signal MC this does not have a big effect (the scaling > factor is close to 1), but it seems nicer to understand it I think. > > Thanks, > Kerstin > > > On Mon, 24 Apr 2006, Kerstin Tackmann wrote: > > > > > Hi Concezio, > > > > > I am not sure I understand your point... the charged/neutral B > > > correction is applied on B+ montecarlo and takes care of the fact that > > > the semi-exclusive reconstruction works differently in data and MC. For > > > instance we don't simulate all the decay modes we reconstruct in data, > > > so we need to rescale the MC charged/neutral B ratios to what we see in > > > data. > > We don't? Not even in the generic BBbar MC? I was not aware of that. > > > > > Since we are correcting for Breco effects, we compute the > > > correction factors at the level of the semileptonic events selection and > > > carry them over the entire analysis afterwards. > > Yes, I saw that. I'll try to be a little more precise about what I mean, > > that was maybe somewhat buried in my lenghty explanation. > > > > My worry is the following: > > > > To compute this correction factor (between charged and neutral B), only the > > right sign sample is used for the neutral B. > > Since there are only requirements on the lepton so far, data is a > > mixture of b->clnu and b->ulnu, probably much b->clnu, since no of the > > veto cuts is applied yet. > > > > Now the thought about the mixing (not the mixing correction) comes in: > > b->clnu has cascades, b->ulnu does not have cascades. Therefore, b->clnu > > has a different ratio of right sign to wrong sign for the neutral B than > > b->ulnu. Is this correct? Therefore, b->clnu has a different ratio in > > right sign neutral B events than charged B events. Is this correct? > > If both these are correct, we expect data (which is dominated by b->clnu > > at this stage of cuts) to have a different ratio of right sign neutral B > > events and charged B events than signal MC (and pure b->ulnu data if we > > had that). Right? > > > > Yet, we use this ratio of b->clnu dominated data to reweight the signal > > MC, although we do not expect the ratio between the two to be the same, > > based on the signal side. > > > > I understand we do this reweighting to correct for things on the tag > > side, but in order to do this, it seems we need to assume that all the > > difference (that we even out) comes from the tag side. Unless I am > > missing something, that assumption does not seem to be justified, since > > we expect differences on the signal side as well, which are physics and > > should not be evened out. > > > > > The mixing correction is applied on neutral B mesons only, and > > > separately on data, signal MC and background MC. It is independent on > > > data/MC yields for charged and neutral Bs. > > Yes, I saw that in VirUtil. > > > > I hope I did a better job explaining what I mean ;-) Does this make more > > sense? > > > > Cheers, > > Kerstin > > > > > > > > > Ciao, Concezio. > > > > > > PS: tomorrow Italy celebrates the end of WWII, so I guess we won't show > > > up for the meeting... > > > > > > Kerstin Tackmann wrote: > > > > > > >Hi, > > > > > > > >sorry, I realized that the email below misex up same and opposite sign > > > >events. We use the ratio between charged events and opposite sign neutral > > > >events, of course. Sorry about that. > > > > > > > >But apart from that, why was it done the way it was? I'd just like > > > >to understand and see if we should be doing the same... Daniele? Concezio? > > > > > > > >Thanks, > > > >Kerstin > > > > > > > > > > > >On Tue, 18 Apr 2006, Kerstin Tackmann wrote: > > > > > > > > > > > > > > > >>Hi all, > > > >> > > > >>I have a question concerning the correction we apply to reweight the MC > > > >>to have the same ratio of charged to neutral B as data. > > > >> > > > >>More precisely, we apply this correction to all types of MC (signal, > > > >>b->clnu, other) and we really only use the same sign events for the > > > >>neutral B decays. > > > >> > > > >>Here is what I wonder about: When we do the mixing correction (see BAD > > > >>540 section 5.2.1), we assume that among the neutral B, all the leptons > > > >>we call signal leptons arise from direct (B->Xclnu) and from cascade > > > >>decays, mixed and unmixed in both cases. In signal MC, we will not have > > > >>any cascade leptons from D decays, so the ratio between same sign and > > > >>opposite sign events will be different in signal MC and in b->clnu MC > > > >>and in data. > > > >>This means that also the ratio between charged B and same sign neutral B > > > >>events will be different in data (which, with the few cuts applied here, > > > >>will be mostly b->clnu and hence has cascade leptons from D) and signal > > > >>MC. Yet we reweight the signal MC to have the same ratio. > > > >>So it seems we take the cascade leptons into account for the mixing > > > >>correction, but not for the correction between charged and (same sign) > > > >>neutral B decays. Was there an explicit decision to do it this way? > > > >>If so, why? Does whoever implemented it this way remember the reasoning > > > >>behind this? > > > >> > > > >>I stumbled over this when I went through the VVF code to see how this > > > >>correction should be applied to the spectral unfolding specific > > > >>data sets. However, this seems relevant for the BRBR extraction as well. > > > >> > > > >>Thanks, > > > >>Kerstin > > > >> > > > >> > > > >> > > > >> > > > > > > > >