Hi all,
as you probably remember, the three shapes fit came always out with
an error 20% lower with respect to the one obtained with background
subtracted method.
We have never understood completely the reason for this. After the
Riccardo's mumblings about this problem and a discussion we had
this afternoon, we realized that the likelihood method we were using
cannot treat properly the non poissionian error in each Mx bin due
to the mes fits (since the likelihood fit does not use that information
at all).
Then we decided to switch to the chi square method and we obtained
compatible results between subtraction and three shapes fits (same error
and central value).
So, we should use chi square technique instead of the likelihood one in
order to get the right error. We propose to switch to the chi square
method. The systematics due to the fact that the chisq does not
weight properly low statistics bins will be calculated changing the binning.
As a consequence, I did not post the results. Let us know if you agree.
Here I show you how the new numbers look like. In general the error will
be a 25% higher than before. All the disagreements in the various
subsamples will be then reduced.
All
---
BRBR = 0.0164 +- 0.0040(stat) +- 0.0012(MC stat)
and this is the plot:
http://www.slac.stanford.edu/~daniele/vub/dataplots/datafitresults_nocat.eps
Red histo is the total. The blue is the b->clnu. The green one is
the "other" component.
Here the results dividing the sample in different subsamples:
B0:
---
BRBR = 0.0032 +- 0.0058(stat) +- 0.0019(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/dataB0fitresults_nocat.eps
Bch:
----
BRBR = 0.0266 +- 0.0054(stat) +- 0.0018(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/dataBchfitresults_nocat.eps
electrons:
----------
fit problems here. Error < 1 in one of the Mx bins. We are fixing these.
muons:
------
BRBR = 0.0249 +- 0.0065(stat) +- 0.0022(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datamufitresults_nocat.eps
run1:
-----
BRBR = 0.0094 +- 0.0055(stat) +- 0.0009(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datarun1fitresults_nocat.eps
run2:
-----
BRBR = 0.0216 +- 0.0055(stat) +- 0.0014(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datarun2fitresults_nocat.eps
superblock1 (Breco int purity>80%)
----------------------------------
BRBR = 0.0145 +- 0.0071(stat) +- 0.0013(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datasb1fitresults_nocat.eps
superblock2 (Breco 80%>int purity>50%)
----------------------------------
BRBR = 0.0175 +- 0.0057(stat) +- 0.0013(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datasb2fitresults_nocat.eps
superblock3 (int Breco purity<50%, single mode purity>10%)
----------------------------------------------------------
BRBR = 0.0187 +- 0.0076(stat) +- 0.0010(MC stat)
http://www.slac.stanford.edu/~daniele/vub/dataplots/datasb3fitresults_nocat.eps
The results on the depleted sample can be found at
http://www.slac.stanford.edu/~daniele/vub/dataplots/depleted.txt
Daniele
|