Hi all (again),
this is a first test of the fit of ratio(BR) on the depleted sample.
The result should not be zero (there is a small amount of b->ulnu events
in the depleted sample and the efficiency are calculated correctly) but
you should get something compatible with the generated ratio(BR)
with huge errors. I fitted also the data distribution since the result
is actually still blinded due to the big error.
Here the results:
COCKTAIL
--------
- ratio(BR) = 0.013 +- 0.012
http://www.slac.stanford.edu/~daniele/vub/newdeplMCfitresults_nocat.eps
* With three shapes (w/o background subtraction)
- ratio(BR) = 0.014 +- 0.012
http://www.slac.stanford.edu/~daniele/vub/newdeplMCshapefitresults_nocat.eps
GENERIC
-------
- ratio(BR) = 0.050 +- 0.022
http://www.slac.stanford.edu/~daniele/vub/newdeplgenefitresults_nocat.eps
* With three shapes (w/o background subtraction)
- ratio(BR) = 0.067 +- 0.021
http://www.slac.stanford.edu/~daniele/vub/newdeplgeneshapefitresults_nocat.eps
DATA
----
- ratio(BR) = 0.09 +- 0.04
http://www.slac.stanford.edu/~daniele/vub/newdepldatafitresults_nocat.eps
* With three shapes (w/o background subtraction)
- ratio(BR) = 0.13 +- 0.03
http://www.slac.stanford.edu/~daniele/vub/newdepldatashapefitresults_nocat.eps
As you see
* cocktail MC is fine
* generic MC is ~ fine: 1-1.5 sigma off
* data are off: 2-3 sigmas off
Looking again at the plot:
http://www.slac.stanford.edu/~daniele/vub/newdepldatashapefitresults_nocat.eps
I have two comments:
1) on data the resolution seems to be different and there is
also a positive bias. I used the root files from the official chains.
I did not apply any additional smearing (I assume PID and track
killing are already there; Urs, may you confrim this?). We can
conclude that the difference is due to track smearing + neutral
smearing + (?)
2) looking at the plot in the BAD the agreement data-generic MC for the
depleted sample seems to be perfect. Since Cocktail and generic MC
seem to be similar (
look at
http://www.slac.stanford.edu/~daniele/vub/newdeplgeneshapefitresults_nocat.eps
the model used in the fit is cocktail)
and since cocktail and data seem to be in disagreement (
look at
http://www.slac.stanford.edu/~daniele/vub/newdepldatashapefitresults_nocat.eps
the model used in the fit is again the cocktail)
this implies that generic and data do not agree. Urs, are you applying
an additional smearing to the generic MC? If not, something strange is
going on here.
Many further tests will be performed including neutral + tracking
smearing in order to understand this discrepancy.
Daniele
|