> > I think the first plots (the mc truth ones) need to be remade, I cannot > > believe them, but if you are really positive about the plot that says that > > basically all candidates we select with like>0 are true Kl, > > that plot is using the information on generic that comes from the MC > matching association: if the information is not reliable the plot cannot > be trusted. > I'm going to redo the plot directly on true (not reconstructed) KL were I > do not need to use the associator and see if the statement is still valid. Too fast. The function that returns the likelihood of course rely on reco information. So I can't call it on true KL. How should I proceed from that? How to ensure ourselves that the likelihood method is selecting true KL? alessio > > > then I would > > also suggest Urs runs with the data-MC comparison you showed the other day > > ciao > > ric > > >