Hi all,
here is a status report.
The summary is still: work in progress :-(
(1) we noted at our last meeting that we were doing mES fits on the data
depleted sample by using the same signal and background PDF parameters
determined on the MC enriched sample. So we re-determine the PDF
parameters to be used on the data depleted sample in the following way.
a) peaking background PDF: fit all parameters on the entire MC-depleted,
MC-truth-unmatched sample after applying all cuts. Keep them fixed when
running on data
b) signal PDF: fit all parameters on the entire MC-depleted, MC-truth-
matched sample after applying all cuts. Then, on single mX bins in data:
if((number of entries for mES>5.27) < 1000.){
take signal parameters parameters from b) and keep them fixed
} else {
fit sigma_L and sigma_r1 (i.e. the asymmetric widths of the signal PDF)
while keeping the other parameters fixed to the values determined in b)
}
The resulting mES fits on the data depleted sample are here:
http://www.slac.stanford.edu/~bozzi/datadepl/
the various eps files correspond to mX bins (the lower and upper limits
are encoded in the filenames, e.g. datadepl1.551.90.eps correspond to
1.55<mX<1.90). There are a couple of bins where we fit 0 peaking
background events, but they are low statistics so they should not
influence the mX fit too much.
Based on these mES fits, we build a new set of signal/peakingBG
corrections. Here is a comparison wrt the old one:
NEW corrections:
#mx_l mx_h corr err_corr
0.00 1.55 1.499 +- 0.495
1.55 1.90 2.688 +- 0.655
1.90 2.20 1.801 +- 0.296
2.20 2.50 1.896 +- 0.611
2.50 2.80 1.165 +- 0.468
2.80 3.10 0.637 +- 0.311
3.10 3.40 19.367+- 34.585
3.40 3.70 1.524 +- 1.610
3.70 4.20 8.180 +- 31.833
4.20 5.00 0.555 +- 6.639
OLD corrections:
#mx_l mx_h corr err_corr
0.00 1.55 2.377944 +- 0.817079
1.55 1.90 5.242231 +- 1.379597
1.90 2.20 2.988877 +- 0.512671
2.20 2.50 2.354238 +- 0.642029
2.50 2.80 1.337182 +- 0.425876
2.80 3.10 1.127306 +- 0.613821
3.10 3.40 1.197736 +- 1.071731
3.40 3.70 1.445228 +- 1.332675
3.70 4.20 1.781268 +- 2.058948
4.20 5.00 0.086513 +- 0.641369
Results for the mX fits are in
http://www.slac.stanford.edu/~bozzi/scra/Ibunewcorr_fixcount/
to be compared with the old correction in
http://www.slac.stanford.edu/~bozzi/scra/Ibutest_fixed_counttag/
Since the correction factors are somewhat lower in the low mX regions,
we fit less Vub events (333 to be compared with 416) and therefore we
get a smaller PBRBR. However, the chisquare of the fit is still high
(8.5, to be compared with 7.8).
(2) We are still working on the parameterization of the "other"
background. 3-parameter mX fits give always 0 for the other component,
no matter whether you fix the S/peakingBG in mES fits or not, or if you
take the enriched or depleted MC sample. Examples are in
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_nounf_fixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_nounf_fixcount/3p_nounf_fixcountfitresults.eps
(S/peakingBG fixed)
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_nounf_nofixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_nounf_nofixcount/3p_nounf_nofixcountfitresults.eps
(S/peakingBG not fixed)
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_depl_nounf_fixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_depl_nounf_fixcount/3p_depl_nounf_fixcountfitresults.eps(S/peakingBG fixed, depleted sample)
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_depl_nounf_nofixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibu3p_depl_nounf_nofixcount/3p_depl_nounf_nofixcountfitresults.eps
(S/peakingBG not fixed, depleted sample)
We therefore resumed an old flag (-fitdss) of VirFit with the following
meaning
-fitdss=1:
vcb component is Dlu+D*lnu
other component is D**lnu + non-SL events
-fitdss=2:
vcb component is Dlnu+D*lnu+ non-SL events
other component is D**lnu only
Results are in
http://www.slac.stanford.edu/~bozzi/scra/Ibufitdss1_nounf_newfixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibufitdss1_nounf_nofixcount/fitdss1_nounf_nofixcountfitresults.eps
(fitdss=1, new S/peakingBG correction ratios)
http://www.slac.stanford.edu/~bozzi/scra/Ibufitdss2_nounf_newfixcount/
http://www.slac.stanford.edu/~bozzi/scra/Ibufitdss2_nounf_newfixcount/fitdss2_nounf_newfixcountfitresults.eps
(fitdss=2, new S/peakingBG correction ratios)
Distributions look better, fit chisquares are 4.3 (fitdss=1) and 5.8
(fitdss=2). This time, most of the chisquare is due to the bin
2.2<mX<2.5.
(3) Let's take fitdss=1. The ratio of other/vcb components is
0.1025/0.1859. If we fix the other/vcb components to this value in the
unfolding mX fit we get
http://www.slac.stanford.edu/~bozzi/scra/Ibunewcorr_fixcount_fixdss/
http://www.slac.stanford.edu/~bozzi/scra/Ibunewcorr_fixcount_fixdss/newcorr_fixcount_fixdssfitresults.eps
chisquare per DOF is about 4, dominated by a ~+4sigma discrepancy in
the bin 2.2<mX<2.5. and a ~-2sigma discrepancy in the bin 1.9<mX<2.2
(hills and ditches compensate...). Please don't look at results other
than the mX fit. The pstarfactor and number of semileptonic events might
be not correct. A new version of the last fit with the correct numbers
will appear in the above link soon.
Any comments?
Concezio and Antonio.
