Hello Folks,
I am trying to implement the new hybrid model in our generator and I am
slightly confused.
I was trying to check that the generator level distribution of Mx after
the generator level reweighting was the same as the reweighted
distribution at VubFit level. I was failing and I traced it back to a
different nonresonant Vub spectrum.

If I compare mxhadgen in
with the corresponding distribution at generaltor level with no cuts
(top=generator; bottom = mxhadgen). The mxhadgen distribution is clearly
harder, its mean being 200 MeV higher.

If I then request the presence of a reconstructed lepton, nle==1, (with
pcms>1) I get the agreement in
which is basically ok.

I investigate the possibility of having problems with the brehmsstrahlung,
and I found that the generator level reweighting ignores possible
bremsstrahlung (i.e. the photon is not assigned to the 'hadronic mass'
while it is in our analysis level reweighting. Looking at electrons and
muons separately this seems to be a tiny effect, though.

 I then looked at a possible bias of the b-reconstruction
and I actually found a strong dependence of the mean value of mxhadgen
with the purity of the mode:
(note that at generator level <mxhad>=1.45). This means that the mes
sidebands are dominated by high mx states, which makes sense because of

In summary:
1) when looking at the generator level mx we need to make sideband
subtraction (this might have been obvious to you, but not to me ...)
Does anybody have a sideband sub distribution of mxhadgen before any cut
so that I can test the generator?
2) are we computing the weights on the sideband subtracted or the raw
events? Are we scaling to the #of generic with or without sideband
subtraction? I would like to double check the logic of this process
because it can be delicate ...
3) I am afraid we have to redo the weights at generator level in order to
be sure , also given the brehmstrahlung business