Hi all,
I made some changes to recoilDSys in RecoilAnalysis last February, but I
did not touch the recoilDSys in VubAnalysis.
I posted a short description of the changes is posted here:
http://babar-hn.slac.stanford.edu:5090/HyperNews/get/semi_lept_decays/4.html
But since it refers to the AWG meeting before the posting, it might be a
little unclear. I made three types of changes to the code:
1) I changed the weights that were used for the B->Xclnu BFs according to
the values of the AWG-reweighting page. Since the uncertainties there were
asymmetric I implemented an asymmetric Gaussian in the Brandomize function
for the variation of the BFs.
2) I had problems with how the random numbers were thrown, since it used
the time to set the seed and when jobs would start too shortly after
another they would give me the same results (which is obviously not what
we want). So I changed it so you could pass a number that determined how
the seed got set (in pratice it throws random numbers for 10*the number
you pass it and uses the random numbers it throws after throwing these
10*x away, and accordingly for the D weights). So you would call
VirVubFitter with -Sys i, i=3...52 or so and were sure to get different
random numbers thrown per job (and I made checks the distributions of the
weights looks ok doing it this way and to not have some strange
behavior).
3) This is for the unfolding, so we could write out the weights into files
and read them in later again when running in our binning.
This version was tagged KT022304. I do not know if Concezio's changes were
made after this.
Kerstin
On Fri, 7 Jan 2005, Concezio Bozzi wrote:
> Hi all,
> in VirVubFitter we have been using RecoilAnalysis/recoilDSys, but we
> also did (more than 1 year ago) detailed checks with
> VubAnalysis/recoilDSys, which I believe gave consistent results.
> At the moment we cannot check the details in our logbooks (both Virginia
> and I are not in Ferrara), we'll do that on Monday.
> Ciao, Concezio.
>
> Daniele del Re wrote:
>
> >>yes, in principle you are right. However, VubAnalysis/fitNtp.cc
> >>includes the other version (the one in VubAnalysis). So it would be
> >>nice to have THE person to know sign off with a better CL than "I
> >>presume". Are the differences between the two versions understood?
> >>
> >>
> >
> >VirVubFitter results were consistent with the fitNtp ones (right
> >Virginia) and fitNtp is not used since ages.
> >
> >This is what I know. I am noticing that the VubAnalysis version was the
> >result of the migration done by Ed. Probably Ed himself knows if there was
> >an original large difference between the two. After the first commit there
> >have been just two versions with fixes by Kerstin and Concezio.
> >
> > Daniele
> >
> >
> >
> >
>
>
|