Hi Wolfgang,
technically speaking, I think that the pstarsample dataset, already
implemented in VVF by Roberto, does what you need. That dataset contains
vcb and vub events from the same ntuples (generic MC), and the vub
events are reweighted as usual, except the magic k factors which are
used to normalize generic MC and signal MC to the same luminosity. So if
you compute the pstarfactor from the pstarsample you should be done.
The trick with the total charmless BF is the following. When you apply
the reweighting to the non-resonant MC you also inflate at the same time
the non-resonant BF in MC, which is a factor ~2 lower than what we see
in data. Indeed, different weight files were produced in the past which
correspond to different values for the total b->ulnu BF, as measured
e.g. in the mX, or the q2-El, or the endpoint analyses. For this reason,
after reweighting non-resonant events and rescaled resonant BFs, the
BF(B->ulnu) which you need in order to obtain the pstarfactor is the one
corresponding to the weight files you are using. Last time we used a set
of weights corresponding to BF(B->ulnu)=0.00224, the PRL mX result.
A small complication arises when doing the unfolding. In this case the
formula for the pstarfactor should have the PBF(B->ulnu), due to the
kinematic cuts. In other words, you should multiply BF(B->ulnu) by the
DFN acceptance. You get the latter from getEffFromDFN in VirUtil in the
1D case, or from getEffFromDFN2D in the 2D case.
Talk to you later, Concezio.
Wolfgang Menges wrote:
>Hi all,
>
> I am a bit stuck with the pstartfactor calculation. Well, it is
>not the actually calculation but the reweighting of the data sample. I
>will outline in this mail the problem and we can discuss it in todays
>meeting.
>
>The p*factor is calculated from generic BB sample. This sample is used for
>the Vcb AND Vub components which avoids to fix (or know) the ratio between
>the two samples. The P*factor is then quite simple:
>
> e^u_t e^u_l N^u_l BF(B->clv)
> --------------- = ------- * ------------
> e^sl_t e^sl_l N^sl_l BF(B->ulv)
>
>The sample has the same luminosity and cross section for B->clv and
>B->ulv. So, that's easy.
>
>Now the difficult part. For the Vub component, we like to apply the usual
>reweighting, that means rescaling the resonant decays and reweighting the
>non-resonant part. Rescaling the resonant decays is straight forward as
>the BF in generic and signal MC are the same. The other part is the tricky
>part.
>
>First, the generic non-resonant part is already reweighted. Here I can
>take the weights from the MC production and reweight the distribution to
>an unweighted one. Then I can apply the standard weights used also for the
>MC signal sample.
>
>Second, how do I fix the total b->ulv BF or how do I get the total b->ulv
>BF? Maybe the later approach is easier. I can sum the new BF for the
>resonant decays. This gives me the exclusive BF. How to I get the
>inclusive BF? Are the weights normalised in such a way that the inclusive
>BF will remain invariant?
>
>Any comments on this?
>
>Cheers,
>
> Wolfgang
>
>-----------------------------------------------------------------------------
>Wolfgang Menges
>Queen Mary, University of London SLAC, MS 35
>Mile End Road 2575 Sand Hill Road
>London, E1 4NS, UK Menlo Park, CA 94025, USA
>++44 20 7882 3753 ++1 650 926 8503
> [log in to unmask]
>-----------------------------------------------------------------------------
>
>
|