 |
 |
Hi Daniele, just on short comment (see below): On Tue, 2 Jul 2002, Daniele del Re wrote: > > Hi all, > > let's look carefully at the fit results with the official cut ch=0 and > the looser cut ch = -1,0,1 > > http://www.slac.stanford.edu/~daniele/vub/off.eps ch = 0 > > http://www.slac.stanford.edu/~daniele/vub/ch101.eps ch = -1,0,1 > > If you look at the subtracted plot (right one) you will see that the two > distributions are pretty similar exept for the forth bin. > > In > > http://www.slac.stanford.edu/~daniele/vub/compch0-ch101.eps > > I overlay the two distributions. The problem in the forth bin is clear > from this plot. The agreement for the remaining plots is nice and the > increase is compatible with the increase you expect in the signal efficiency > (~25%). > > I ran also the Mx scan with the cut ch = -1,0,1 > > scan > > http://www.slac.stanford.edu/~daniele/vub/newmxstudy.ps > > difference with uncorr error (default is 1.55GeV) > > http://www.slac.stanford.edu/~daniele/vub/compmxstudy.ps > > > The behaviour seems to be pretty unstable. At 1.2 the difference with the > default is 2 sigmas. I would like to point out that with a harder cut on > Mx the results with ch = -1,0,1 and ch = 0 are very compatible. > > We could conclude that the different value we get with ch = -1,0,1 is a > consequence of a wrong background estimation close to the D mass and not > of a wrong signal efficiency. Since the result is very unstable as a > function of Mx this wrong bkg estimation is in the ch = -1,1 sample. naively one would expect that a softer cut in DELTAQ would lead to less model dependence (e.g. in the limit of no cut on DELTAQ you do not dependent on the modeling of DELTAQ distribution from the MC at all). Or in other words, claiming you have a MC-DATA discrepancy for events with DELTAQ=+-1 BUT not for DELTAQ=0 is hard to believe and seems to me unlikely ... someone might even say that your statement is a contradiction by it self .... . > > Daniele > > > > > > > > >