Hi Kerstin,
in general I would say that the situation is really good, you have small
differences. I suspect that you have some difference in the PID
killing and in particular in the seed used to extract the random number to
perform it. I think that Urs can comment on this.
BTW let's go trough your posting.
> for the comparison. The plots using your reduced ntuples for the Vub
> signal:
>
> http://www.slac.stanford.edu/~kerstin/vubfit/s00_mxresult.eps
>
> and the plots using our reduced ntuples for Vub signal:
>
> http://www.slac.stanford.edu/~kerstin/vubfit/all_mxresult.eps
>
> Should we worry because of the small differences?
>
since you have only reproduced the vub component, I don't think that
looking at the those plots is useful. Some numbers in the fit output .dat
files can be very interesting, like the efficiency ones. For instances I
see:
our
EFFICIENCY Vub
Vub total MC (lepton cut) = 7209.87
Vub MC (all cuts) = 2320.44
Vub MC (all cuts + Mx cut) = 1809.65
Vub gene total MC (lepton cut) = 7209.87
Vub gene MC (all cuts) = 2349.14
Eps_u = 0.325823 +- 0.00551968
Eps_Mx = 0.770345 +- 0.00867813
Eps_tot = 0.250996 +- 0.00510637
your
EFFICIENCY Vub
Vub total MC (lepton cut) = 7155.2
Vub MC (all cuts) = 2317.06
Vub MC (all cuts + Mx cut) = 1798.27
Vub gene total MC (lepton cut) = 7155.2
Vub gene MC (all cuts) = 2340.47
Eps_u = 0.3271 +- 0.00554631
Eps_Mx = 0.76834 +- 0.0087207
Eps_tot = 0.251324 +- 0.00512806
here I see differences<<1%, that's good.
> We also compared the files (again the Vub signal MC) in a few variables
> and here are the results (all the three years together):
>
> NONRES
> ******
> your files our files
> # Breco candidates: 202442 202496
> # Breco cand. with mes>5.27 24365 24358
> # events with pcms>1. 6157 6097
> # events with |Gvxbtyp|==7 194735 194784
> # events with |Gvxbtyp|!=7 7707 7712
>
> MIX
> ***
> your files our files
> # Breco candidates: 297479 296048
> # Breco cand. with mes>5.27 36076 35757
> # events with pcms>1. 8976 8839
> # events with |Gvxbtyp|==7 182942 182986
> # events with |Gvxbtyp|!=7 114537 113062
> # events with kplus>0. 190560 190629
>
> The two most important things might be:
> 1. We are seeing some more Breco candidates, but noticeably less leptons.
> What could be the reason for this?
> 2. There are differences in Gvxbtyp, which, to my understanding, only
> depends on truth information of the events. How can these differ between
> your and our files?
how those numbers compare with the fit results? It looks like they are
worse that the fit ones. They are not a result of an mes fit (I suppose),
then I would say that the differences are all in the combinatorics (that
is consistent with a removal of BAD breco modes).
another point. Which is the option "-d" you are using? -4 or -8. If you
have -4 you will pick up only events with a lepton. Even the number
of events with |Gvxbtyp|==7 can depend on this.
>
> Ric mentioned earlier that small differences can be due to minor changes
> in the (Breco) code. But we think Alessio said that the instructions page
> for IslBrecoilUser has the tags that you actually used when producing your
> ntuples. We used the tags given there as well. If we used the same tags,
> how can we get differences in the ntuples?
>
> One more question about the Gvxbtyp variable as it is filled in
> mcTruth.icc:
> If two semileptonic B decays are present in one event Gvxbtyp is set to
> the value that corresponds to one of these decays while the information
> about the other one is lost for this variable (e.g. in your nonres file,
> there are events for which this variable is not 7 or -7 although each of
> the events contains one b B -> X_u l nu decay). Why are you doing it this
> way?
in general, when you have two semiletponic decays, you will not have any
peak in mes. Then, if you extract your yields by mes fit, those cases are
not interesting at all (they will be subtracted).
Daniele
|
