 |
 |
Hi Oliver, I have run over our root files checking how often a track which is MC matched with a pion (TMath::Abs(idTrk[it])==211) is also reconstructed as an electron with the likelihood selector: if(elecIdTrk[it] & 32 ) I have also requested the track to be in the acceptance (0.41<theta<2.37 rad) , but this does not make a large difference. I have tried 2000 and 2002a separately. What I get is in http://www.slac.stanford.edu/~rfaccini/phys/vub/deltam/mis2002a.eps So the misid's range between 1 and 4 % which are an order of magnitude smaller than what you say. Can you try and produce the sample plot (i.e. with the same procedure)? This said these numbers are at least an order of magnitude larger than what the PID BADs declare and that are actually written in the pid tables. This might also explain why we see percent adjustments in our results if we turn on pidkilling at root level. I will investigate a bit more, but this might go in the direction of having to use the tables on root instead of using the PID as declared by Beta. Did you try applying the pid tables at hbook level? ciao ric On Thu, 20 Mar 2003, Oliver Buchmueller wrote: > > Dear All, > > attached you will find a plot showing a comparison > of the number of electron candidates per event for DATA > and Monte Carlo. There is a clear shift in the MC towards higher > multiplicities. At least that seems to be consistent with > our finding of a higher mis-id rate in the MC. > Therefore, requiring only one electron in the event > will lead to a wrong efficiency estimate in the MC. > Again, it remains unclear whether or not this is a bug > in the hbook ntuples (base for the root ntuples) or just > in our code. Concerning the later, Henning and I are still > investigating but no sign of a bug yet! > > Greetings from the America empire. > > Oliver > > > > On Thu, 20 Mar 2003, Oliver Buchmueller wrote: > > > > > > > Dear Breco_root_users, > > > > In our attempt to understand our background problem > > in the hadronic analysis Henning and I made the "discovery" > > that roughly 40% of our total B electron background stems > > from fake electrons (mostly pions). Since that number seems > > to be far off scale we checked the electron mis-id rate for > > various particles. Attached to this mail you will find a plot > > showing these mis-id rates obtained from the generic hbook ntuples. > > This plot also shows a comparison of the mis-id rates for > > clv and non clv events. > > > > Looking at this numbers you will conclude that for example > > the mis-id for pions is one order of magnitude larger that it > > is supposed to be and also the momentum dependence seems not > > to be correct. Although Henning and myself checked our code > > very carefully there is always a chance of a bug. > > We, therefore, would really appreciate if one of you could > > cross check this findings by utilizing your root-tuples. > > For simplicity a test of the pion mis-id would be enough. > > > > If this effect turns out to be real (still hope that Henning > > and I are to stupid to calculate the mis-id rates) quit a few > > analyses might be affected. Anyway, whatever the outcome might be > > this test will at least help us to understand our background > > problem (either we have a bug in our code or the ntuples are > > corrupted). > > > > Thank you very much for your help! > > > > Oliver and Henning >