Heiko Lacker ha scritto: > Hi, > > On Wed, 5 Jul 2006, Antonio Petrella wrote: > >> Hi all, >> >> today I've run again two fit tests, one with the ratio signal/peaking >> bkg fixed (only on data, this time) and one without fixing the ratio. >> >> Results can be found here: >> >> http://www.slac.stanford.edu/~petrella/tmp/scra/Ibutest_fixed/ >> http://www.slac.stanford.edu/~petrella/tmp/scra/Ibutest_SUN/ >> >> General comments: >> >> -comparing the right lower plot (which contains the sum of Bch and B0 >> for data) >> >> http://www.slac.stanford.edu/~petrella/tmp/scra/Ibutest_fixed/test_fixeddatachop_all-1.eps >> http://www.slac.stanford.edu/~petrella/tmp/scra/Ibutest_SUN/test_SUNdatachop_all-1.eps >> >> we see that the effect of fixing the ratio is to increase the >> distribution in the low mx region ( 1.55 < mx < 2.1) and to decrease in >> the high mx region (mx > 2.8 ). >> >> So we looked with more attention to fits performed bin by bin to compute >> the correction factor and...: >> >> - the fit in the high mx region for MC depleted has peaking background >> consistent with 0 >> http://www.slac.stanford.edu/~petrella/tmp/AWG38/depleted/MCsameEND/MCfit_1D_3.103.40.eps >> >> http://www.slac.stanford.edu/~petrella/tmp/AWG38/depleted/MCsameEND/MCfit_1D_3.403.70.eps >> >> moreover in the Enriched MC the signal and the peaking components are >> swapped for 3.1 < mx < 3.4! > This bin is indeed special in the sense that the fit error on the ratio > sig/peakBG is much smaller than the ones in the neighbourhood and that > the ratio has a minimum there. I would think that an average over a larger > mX range would be more appropriate. The question is how to define these > larger ranges. > >> http://www.slac.stanford.edu/~petrella/tmp/AWG38/enriched/MC/MC_enrich_1D_3.103.40.eps >> >> >> - in the low mx region the increase in the mx distribution is due to the >> behaviour of the double ratio (sig/peak)_DATA/(sig/peak)_MC, where the >> numerator and denominator go in the opposite direction i.e.: for MC the >> ratio is small while for data is high, so the net effect of the double >> ratio is to give a high correction. >> >> But fits on MC present clearly some problems: look at >> >> http://www.slac.stanford.edu/~petrella/tmp/AWG38/depleted/MCsameEND/MCfit_1D_1.551.90.eps >> http://www.slac.stanford.edu/~petrella/tmp/AWG38/depleted/MCsameEND/MCfit_1D_1.902.20.eps > Do these fits correspond to to the two bins where we have the larger > correction from the double-ratio? Yes, they are. Here are the actual correction factors: mx bin S/P 0-1.55 3.279662 +- 1.171573 1.55-1.90 9.944666 +- 2.600361 1.90-2.20 6.860380 +- 1.603589 2.20-2.50 1.507182 +- 0.594469 2.50-2.80 0.741645 +- 0.377487 2.80-3.10 0.703099 +- 0.561601 3.10-3.40 0.014759 +- 0.076725 3.40-3.70 0.012368 +- 0.230214 3.70-4.20 1.364814 +- 2.520936 4.20-5.00 3.378136 +- 13.007195 Antonio > > Heiko > >> So I don't think we can trust 100% the numbers used to compute the >> correction so far. >> >> Antonio >>