Hi,
I ran the fit using the new 1-d as implemented in VubAnalysis ale_030609.
For the results on data I got consistent numbers with respect to the
Alessio's ones (not exaclty the same because I am using the default p00
root files). The Ric 1-d reweighting gave a shift up of 1.5% as for
Alessio.
Then I looked again at the true Mx distribution and how the resonant and
the non-resonant are merged.
These are the plots from the fit that show the two components after
reweighting.
Default (prl)
http://www.slac.stanford.edu/~daniele/vub/dataFINALdefboth.eps
Ric-1d
http://www.slac.stanford.edu/~daniele/vub/checkboth.eps
There is a region of big disagreement between 2.5 and 3. I discovered
that is due to a small bug in the new code. In that range the weight was set to
the one for the low mass region (that is very small)
Fixing this I got a new default and new plots
new ric 1-d
BRBR = 0.0211058 +- 0.00257061(stat) +- 0.000956233(MC stat)
and plots
http://www.slac.stanford.edu/~daniele/vub/dataFINALdef_ricweightboth.eps
http://www.slac.stanford.edu/~daniele/vub/dataFINALdef_ricweightintegrals.eps
unfortutely this fix enhanced the disagreement and now the shift is more
than 2.5%.
I studied a bit more in detail the reweighted non-resonant component.
This is the comparison plot
http://www.slac.stanford.edu/~daniele/vub/twonres.eps
red is the default and black is the new one. The agreement is good above
1.5 GeV but below seems to be 10-20% off.
This result is somehow consistent with the numbers in the
Alessio's page http://www.slac.stanford.edu/~asarti/recoil/talks/090603.html (mean
weights that show a disagreeement only in the very first bin) but not with
the plot shown by Alessio
(http://www.slac.stanford.edu/~asarti/recoil/mxq2/TestRicgen.eps) in which
I can't see a 20% disagreement below 1.2 GeV.
Most of the difference between the two methods is below 1.5GeV. I have
the impression that this could be possible only if the breco
efficiency is very different between resonant and not-resonant while the
distortion of the non-resonant model is not large enough to create such an
effect.
We should check this.
Daniele
|