Hi Concezio, first of all: congratulations to the new world champion! Second: Very good progress! This looks like a real step forward. Cheers, Heiko On Fri, 7 Jul 2006, Concezio Bozzi wrote: > Hi, > here are some results for the double ratio > > (signal/peakingBG)_MCenriched > ----------------------------- > (signal/peakingBG)_MCdepleted > > > as a function of mx. We use this double ratio, multiplied by the ratio > (signal/peakingBG) in the depleted data sample, in order to fix the > (signal/peakingBG) components in the mES fits on data after all cuts as > a function of mx. > > The various signal (S) and peakingBG (P) components of this double ratio > on MC can be computed on either: > > (1) the entire MC sample, by performing a 3-parameter fit to S, P and > combinatorialBG (B) > > or > > (2) on separate "signal" (reconstructed Breco mode == true Breco mode) > and "background" (reconstructed Breco mode != true Breco mode) samples. > In this case, the signal sample is used to determine S, whereas the > background sample is used to determine P and B. > > Method (2) has a number of advantages on method (1): > - the uncertainty on S and P is smaller > - the bias on S is reduced, since we don't risk to mix up/swap signal > and peaking background due to the very similar functional forms of their > PDFs. > > The spreadsheet attached to this message proves the above two > statements. Excellent! > The upper part of the spreadsheet contains the results obtained with > method (1), the lower part refers to method (2). Look e.g. at the > errors on the peaking background numbers on the enriched sample on > column E5-E14 for method (1) and E18-E27 for method (2). The same is > true on the depleted sample (column Q) and for the signal yields > (columns K and W). > > The bias in method (1) on the number of fitted signal and background > events (columns I5-I14, U5-U14 for signal, AH5-AH14, AI5-AI14 for > background) is also evident by comparing them to the number of expected > events in the datasets (respectively: columns AI18-AI27, AI31-AI40 for > signal, AH18-AH27, AH31-AH40 for background). > > The double ratios which come out are flat as a function of mx and are > compatible with 1, see the light blue cells in the spreadsheet, or the > attached eps file (The number for the last bin in mx is not meaningful > since there are very few events in the mES plot). > > The bottom line is that the signal/peaking background in the mES fitss > can be fixed in our data to the values which we obtain on the depleted > sample, times a MC correction which turns to be about 1 within 10% and a > ~10% uncertainty, which gets higher at high mx. > > This is not the end of the story, of course. The study needs to be > repeated on the MC samples which we use in VVF (vcb+other, vubIN, > voubOUT), since we have to fix the signal/peaking background components > in there as well... > > Stay tuned, Antonio&Concezio. >