 |
 |
more on this item. Here you find the comparison plots of the background subtracted distributions (the plot onthe right you find in the usual ps file for the fit result). In these plots: - top left is B0+Bch: - top right is B0: - bottom left is Bch: ** COMPARISON Q0 - Q1, cocktail as model: http://www.slac.stanford.edu/~daniele/vub/nicecompold.ps ** COMPARISON Q0 - Q1, generic+crossfeed as model: http://www.slac.stanford.edu/~daniele/vub/nicecomp.ps from these two you can learn that the comparison improve when you use the generic MC (look at B0's) ** COMPARISON Q0, cocktail - generic+crossfeed as model: http://www.slac.stanford.edu/~daniele/vub/nicecompcockgenech0.ps ** COMPARISON Q0, cocktail - generic+crossfeed as model: http://www.slac.stanford.edu/~daniele/vub/nicecompcockgenech101.ps here you see that for Q0 the result of the subtraction is basically unchanged. Large differences for Q1 (B0's). I also refitted the sample Abs(Q)==1 (uncorrelated sample with repect to Q=0) on the using gene+crossfeed MC as a model: B0 BRBR = 0.0265976 +- 0.0235882(stat) +- 0.0129815(MC stat) Bch BRBR = 0.0444071 +- 0.0240263(stat) +- 0.0115915(MC stat) All BRBR = 0.0333972 +- 0.0168482(stat) +- 0.0084564(MC stat) then B0 1 sigma up Bch < .8 sigma up total 1 sigma up before (cocktail as a model) was more then two sigmas. I start being convinced that we have no problem here. Daniele On Sun, 7 Jul 2002, Daniele del Re wrote: > > > Hi all, > > finally something that has a large difference between Q0 and Q1. > I did again the test using the generic MC and generic+crossfeed as a > model. > I discovered that the difference we observed between generic MC and > cocktail 0.0169 -> 0.0140 was DUE TO THE RATIO eps(u)sl/eps(c)sl > calcualted in the fit that was affected by a fit on a very dirt > sample (generic MC with no cut applied). > This introduces a large bias in this factor when we fitted the > generic MC (~20% up since the signal before the lepton cut is > overestimated for b->clnu) > > Then these are the results (correcting this effect): > > > generic MC: > ========== > > Q0 > -- > > B0 BRBR = 0.00383 +- 0.00703(stat) > Bch BRBR = 0.02544 +- 0.00541(stat) > All BRBR = 0.01649 +- 0.00428(stat) > > Q1 > -- > > B0 BRBR = 0.01244 +- 0.00773(stat) > Bch BRBR = 0.03057 +- 0.00635(stat) > All BRBR = 0.02277 +- 0.00487(stat) > > > generic + crossfeed MC > ====================== > > Q0 > -- > > B0 BRBR = 0.00290 +- 0.00705(stat) > Bch BRBR = 0.02644 +- 0.00539(stat) > All BRBR = 0.01700 +- 0.00429(stat) > > Q1 > -- > > B0 BRBR = 0.00655 +- 0.00789(stat) > Bch BRBR = 0.03102 +- 0.00635(stat) > All BRBR = 0.02054 +- 0.00492(stat) > > > The usage of generic + crossfeed MC as a very large impact on Q1!!! > especially on B0s. > > > I summarize > > Q0 default gene + crossfeed > -- > > B0 > BRBR 0.0029+-0.0071 0.0024+-0.0071 > > chi2 0.90232 1.24949 > > Bch > BRBR 0.0264+-0.0054 0.0264+-0.0054 > > chi2 1.37897 1.72423 > > All > BRBR 0.0170+-0.0043 0.0170+-0.0043 > > chi2 1.03306 1.56583 > > > > Q1 default gene + crossfeed > -- > > B0 > BRBR 0.00142+-0.0076 0.0066+-0.0079 > > chi2 1.04791 1.56062 > > Bch > > BRBR 0.0336+-0.0063 0.0310+-0.0064 > > chi2 2.58364 1.9846 > > All > BRBR 0.0253+-0.0048 0.02054+-0.0049 > > chi2 1.56594 0.866931 > > > > The disagreement Q0 - Q1 now becomes (using the difference in > quadrature): > > B0 -1.2 sigmas > > Bch -1.3 sigmas > > > I don't quote the difference in the total (B0+Bch) since it could be > wrong due to a not correct ratio B0/Bch (BTW is -1.5 sigmas) > > > Let's think about this very carefully. > > > Daniele > > > > > >