Hi, > I agree we cannot correct SP5 to look like SP6. However, by determining > separate correction factors we definitely get a better data-MC agreement > and get confidence that the fit is working OK. Here are the errors on > the correction factors: > > Entire sample (Run1-4, SP5+6): > > Chi Square of the Fit = 4.48376 > NDOF = 7 > ratio other/vcb = 0.60 +- 0.07 > Vcb comp = 0.188267 +- 0.00631457 > Oth comp = 0.113042 +- 0.0125698 > > > This is what Kerstin got on separate samples: > > > > > **Fit on Run 1+2, SP5: > > > > chi2/ndof = 1.64116 (for 7 dof) > > > > ratio other/vcb = 0.296045 / 0.287339 = 1.03 +- 0.17 > > > > (with errors: > > > > Vcb comp = 0.287339 +- 0.0183957 > > > > Oth comp = 0.296045 +- 0.0442523) > > > > > > > > **Fit on Run 4, SP6: > > > > chi2/ndof = 1.90748 (for 7 dof) > > > > ratio other/vcb = 0.117799 / 0.141224 = 0.83 +- 0.13 > > > > (with errors: > > > > Vcb comp = 0.141224 +- 0.00785652 > > > > Oth comp = 0.117799 +- 0.0174853) > > > A couple of warnings > > 1) I previously run the fits WITHOUT the B&D reweighting (-b -d flags), > so I run them again and the ratio on the entire sample went from 0.55 to > 0.60. I hope Kerstin was more clever than me and run with -b and -d ;-) I did run with -d -b. > 2) Wolfgang mentions that the signal has been reweigthed to match the > Run1+2/Run3+4 luminosity ratios ub data. A point related to this is: > what happens if you run on Run1+2 or Run4 only? As long as the magic k numbers are correct, applying the weights to SP5 when running on Run1+2 should not hurt I think, as it only changes the effective number of events (as most of the other reweightings do), but it keeps the relative fraction as we want them. > In this case you should > not apply any reweighting at all and you should also use different magic > factors. Did anybody check this? I did recompute the magic k numbers when we introduced this SP5/SP6 correction. I am sure I did it for the full statistics and I just crosschecked with my notes that I also recomputed the numbers for the individual run periods as they are in wfermifile.dat, assuming the SP5 vs. SP6 reweighting will be done independent of the run flag. So I think things are used consistently here. Cheers, Kerstin > A similar reweighting should be done for vcb+other background, I guess. > > Ciao, Concezio. > > > Il giorno gio, 20-07-2006 alle 07:40 +0000, Wolfgang Menges ha scritto: > > Hi Kerstin, > > > > Kerstin Tackmann wrote: > > > Hi Concezio, > > > > > >> if we stay with SP5/6 I think that reweighting the D** wrt D+D* > > >> separately for SP5 and SP6 would be the best we can do. But this means > > >> that we need to know whether we are analysing SP5 or SP6 events when we > > >> build the datasets in VVF, which I am not sure how to do. > > >> Using SP8 means a lot of extra work/checks/tunings. > > >> However it is not clear to me why when we compute the D** reweighting we > > >> get 1 for SP5 and 0.83 for SP6, whereas we get 0.55 for (sp5+sp6). Well, > > >> the fit might be just readjusting itself... > > > Would we actually want to trust it if it is readjusting by so much? > > > > > > Well, checking with data would be better but we have to rely on MC here. > > One thing nobody meantioned is the error on this ratios or the ratio > > of the D** components to the total background. > > > > And you can see clearly from the numbers that something else is going > > on. Otherwise you will not end up with (1+0.83)/0.55. Or is this to naive? > > > > Cheers, > > > > Wolfgang > > > >