Hi Antonio, one question I forgot to ask: why did you rescale in the non-extended chi^2 fit with the chi^2 fit you find in the histo of the extended NLL? Cheers, Heiko On Mon, 8 May 2006, Antonio Petrella wrote: > Hello, > > I've posted the chi^2 scaling also for data. > In this case the fit seems to be more unstable: multiplying errors by > 0.1 led to a real bad fit (Chi^2 ~ 1800). > > Heiko Lacker ha scritto: > > >>> 2) I'm a bit surprised when comparing your fit on data with the plot shown > >>> at April 19: The statistics is not exactly the same (order 20% difference). > >>> Why is that? > >> I think it is due to the different tuning of parameters. Looking at the > >> plot of April 19th we can see that the amount of event fitted by > >> crystall ball in the signal region is greater than on the last plot. So > >> the number of signal events is lower. > >> In addition for the fit on data of april 19th I set the endpoint for > >> Argus and Cristall ball fixed, which is not so good when fitting data, > >> is that right? > >> So I would say that the latest fit (5 may) has more correct assumptions. > > This is not exactly what I meant. The number of events in the mES peak > > is different between both plots. > > > > I think it is because the fit of 19 apr. gives more fraction to Cristal > ball than to signal function wrt the fit on 5 may. > > > Antonio >