 |
 |
Hi Daniele, On Thu, 16 May 2002, Daniele del Re wrote: > > Hi all, > > I produced a new page with the comparison data-mc with the correct > sideband subtraction (in each bin of the plotted variable). > > http://www.slac.stanford.edu/~daniele/vub/complots/comp.html > > This is my first set of comments: > > 1) in general the agreement is better in the generic MC with respect to > to Cocktail. > I would rephrase this statement to: "in general the agreement between data and Monte Carlo is bad and it is hard to tell which of the two types of MC performs worse." ... but thats just my personal point of view. > 2) the disagreement is the charged and neutral multiplicities. In > particular the neutral multiplicity seems to be a bit worse in the > high-energy case (160MeV<E<320MeV, E>320MeV). This affects mm2 and mx > distributions. > > 3) if we normalize using the number of reconstructed B's we observe that > cocktail and generic MC's go in different direction (don't look at the > chi^2 in these plots since it is wrong). This effect is also > reflected in the efficiency comparisons. We must study in detail > this point. I am not sure that I really understand what "normalize using the number of reconstructed B's" means. Looking at the plots for MXHAD (allcuts) it turns out that "cocktail vs data" and "generic vs data" are quite similar. I assume that for "allcuts" you normalize the histograms to the same area (correct?). In case of the "allcuts(breconorm)" the normalisation for the cocktail goes up whereas for generic is goes down with respect to the "allcuts" scenario. Thats the effect of "normalize using the number of reconstructed B's" (I guess) ..... how do you get this significant different normalization factors for generic and cocktail. I always thought the normalisation is determined from data and hence the same for cocktail and generic?! Regrads, Oliver > > Daniele > >