 |
 |
Hi Virginia, thanks for the long summary! > _ 2 pdf? (Antonio ) > pbl: this has been seen as the solution to the not converging 3 pdf fitter, > BUT > the Thorsten parametrization still have a kink > > (see for ex. > bkg & signal fitted together - MC: > ~azzolini/Semilep01/fitter-24/VirVubFitter/fittest/allMC_enr_SL_haspi0-1.intp0.00_-10.0010.00.eps > > so we tried to fit only signal: > ~azzolini/Semilep01/fitter-24/VirVubFitter/fittest/sig_enr_SL_haspi0-1.intp0.00_-10.0010.00.eps (A) > > and only bkg: > ~azzolini/Semilep01/fitter-24/VirVubFitter/fittest/bkg_enr_SL_haspi0-1.intp0.00_-10.0010.00.eps (B) > > From (A) we learned that the X^2 is dominated by the kink presence > --> solution: _ we could request the continuity of the first derivative > _ abandon Thorsten signal parametrization, and come back to the origins. > that would mean 2 pdf ( CB + Argus) like in the past but > leaving floating all the parameters, > and not fix some of them as it turned out we were doing in the past. > > For do that we looked at (B), making ourself aware that the way how the signal & bkg are now defined > is no more adapt. :( > Infact, their definion, made through the comparation of modeB reconstructed and thruth mode B, > is not anymore helpful if we decide in this sense. > We should instead def/distinguish signal and bkg and fit them together. Do you have plots from these tests that you could point us to? Most interesting would be to see where you looked at the definition of signal and background and their shapes. Did you do this on R14 or higher? Maybe we should repeat this on signal MC comparing R14 with a higher rel, that should show us how much this depends on the neutrals truth matching bug (which has been fixed in R18 and R20 just for signal MC). I already told Roberto, you can find the R20 skimmed signal MC ntuples here in case these are useful for your tests: /nfs/farm/babar/AWG23/ISL/kerstin/TrkFixup_signal/kerstin/rootples/ (with VRU in 20.2.0d and ChargedTracks+V0BtaTrack for the charged particles on the signal side, skimmed with TrkFixup). Cheers, Kerstin