Hi, 

due to further implementations/testings, we are afraid we cannot provide
a tag for VVF by today. However, we got some interesting results by
running some Mx fits with the following configurations: 

(1) fix the S/peakingBG ratio after applying all cuts: in data by using
the double ratio technique, in MC by fitting separately truth-matched
for S and non-truth-matched events for peakingBG, see 

http://www.slac.stanford.edu/~bozzi/scra/Ibutest_fixed_usecb/

(2) fix the S/peakingBG ratio in data after applying all cuts as in (1),
take all MC quantities (Mx shapes, efficiencies) by simply counting
events surviving the cuts in truth-matched events, see

http://www.slac.stanford.edu/~bozzi/scra/Ibutest_fixed_count/

If our mES fits are OK, the results obtained in (2) should be compatible
with the results obtained in (1). For what the MC shapes are concerned,
you will find comparisons in 

http://www.slac.stanford.edu/~bozzi/comp_vcboth.eps (vcb+other)
http://www.slac.stanford.edu/~bozzi/comp_vubin.eps (vubIN)
http://www.slac.stanford.edu/~bozzi/comp_vubout.eps (VubOUT)

where the points refer to (1), the histograms to (2) and the four plots
are: (top left) chargedB, (top right) neutralB opposite flavor, (bottom
left) neutralB same flavor, (bottom right) sum of the previous ones,
corrected for mixing and charged/neutralB ratio observed in data. 
The Mx distributions obtained with the two approaches are quite
compatible, and indeed the Mx fit on data gives

(1) 
Vub IN  fitted 1' bin = 415.428 +- 1.58277(stat MC) +- 30.8589(err fit)
Vub OUT fitted 1' bin = 41.1093 +- 2.86677(stat MC) +- 3.02161(err fit)
Vcb+Oth fitted 1' bin = 380.874 +- 10.1103(stat MC) +- 11.2604(err fit)

(2)
Vub IN  fitted 1' bin = 415.912 +- 8.34177(stat MC) +- 30.8209(err fit)
Vub OUT fitted 1' bin = 38.0393 +- 2.4665(stat MC) +- 2.85527(err fit)
Vcb+Oth fitted 1' bin = 383.46 +- 8.57534(stat MC) +- 11.0408(err fit)

The efficiencies however are a bit different:

(1) 
Eps_u =  0.440843 +- 0.00529543
Eps_u (mix corr)= 0.445315 +- 0.0053009
Eps_Cut = 0.729764 +- 0.00713367
Eps_PhSp = 0.491828 +- 0.000818
Eps_tot = 0.158227 +- 0.00246454


(2)
Eps_u =  0.408981 +- 0.00537718
Eps_u (mix corr)= 0.40782 +- 0.00537481
Eps_Cut = 0.88187 +- 0.00551991
Eps_PhSp = 0.491828 +- 0.000818
Eps_tot = 0.177387 +- 0.00259984

The last number which enters in the BRBR determination is the pstarfactor:

(1)
pstarfact           1.20276 +- 0.0290694

(2)
pstarfact           1.10181 +- 0.0155489


This counterbalances the different efficiencies, giving quite similar
results for BRBR: 

(1)
PartialBRBR = 0.0133692 +- 0.000993091(stat) +- 0.000418926(MC stat)
BRBR = 0.0271826 +- 0.00201918(stat) +- 0.000852973(MC stat)

(2)
PartialBRBR = 0.0131184 +- 0.00097213(stat) +- 0.000360179(MC stat)
BRBR = 0.0266727 +- 0.00197657(stat) +- 0.00073367(MC stat)

Conclusion: the bias on mES fits on MC obtained after determining the 
S/peakingBG ratio in matched/unmatched events after all cuts is small. 
There is however some bias on the mES fits on bigger samples, namely the 
ones used to compute the efficiencies and the pstarfactor, where we don't 
use matched/unmatched events (e.g. when we require lepton cuts). 
Therefore we propose to count matched events to compute all MC-related 
quantities. BTW, this avoids many mES fits and speeds up VVF quite a lot! 

(3) The above figures can also be compared with the results of VVF without 
applying any S/peakingBG constraint, as in our first attempts, see 

http://www.slac.stanford.edu/~bozzi/scra/Ibutest_nofixed/

Comparisons of Mx MC shapes for (2) and (3) are in 

http://www.slac.stanford.edu/~bozzi/comp_vcboth_nofix.eps
http://www.slac.stanford.edu/~bozzi/comp_vubin_nofix.eps
http://www.slac.stanford.edu/~bozzi/comp_vubout_nofix.eps

points are (3), histograms are (2). Difference are bigger in this case. 
The Mx fit, efficiencies, pstarfactor, BRBR for (3): 

(3)
Vub IN  fitted 1' bin = 494.356 +- 8.76092(stat MC) +- 64.6291(err fit)
Vub OUT fitted 1' bin = 44.0603 +- 4.46821(stat MC) +- 5.65902(err fit)
Vcb+Oth fitted 1' bin = 334.964 +- 16.6672(stat MC) +- 22.8977(err fit)

Eps_u =  0.440843 +- 0.00529543
Eps_u (mix corr)= 0.445315 +- 0.0053009
Eps_Cut = 0.834418 +- 0.00597102
Eps_PhSp = 0.491828 +- 0.000818
Eps_tot = 0.180918 +- 0.00254743

pstarfact           1.20276 +- 0.0290694

PartialBRBR = 0.0139138 +- 0.00181901(stat) +- 0.000577261(MC stat)
BRBR = 0.02829 +- 0.00369847(stat) +- 0.00117465(MC stat)

Note that the statistical error on the fitted Vub events, and therefore on 
PartialBRBR and BRBR are about a factor 2 worse than in (1) and (2). 
This is expected since we are performing mES fits where we determine 
simultaneously S, peakingBG and combinatorialBG without any further 
constraints. 

More news (and new VVF tag!) by tomorrow, 
Ciao, Antonio & Concezio.