 |
 |
and here is the 'appended' material ... Cheers, --U. ---------------------------------------------------------------------- Dear review committee, many thanks for your continued careful reading of our manuscript. We appreciate your scrutiny. > No semileptonic analysis as far as we are aware of is using muons so far. Interesting that you ask this. We are using leptons only in a measurement of a ratio of branching fractions, not for an absolute measurement. As far as we know, there are other efforts in BABAR to determine Vub trying to make use of muons on an absolute scale. We'll have a look at their BAD. > What are the data/MC comparisons which validate the use of muons. > In the muon plot shown in (Figure 14, page 25), > there is a large disagreement between data and MC for the muon spectrum. This *might* be due to the fact that in the previous processing batch (on which the current content of the BAD is based), a uniform set of PidTables corresponding to Run2 was used. The data reflect a mixture of Run1 and Run2. MTC. > We would like to see more comparison plots between data an MC for muons > (theta_lab, theta_B, phi, ... ) We will add comparison plots of data/MC muon kinematics. > How will a systematic error be estimated for the disagreement in Figure 14? With a variation of the PID killing. > In Figures 14 to 18 we would like you to show the different > components in the Monte Carlo, e.g. signal, b->clv, cascade, fakes Will do for next version > What is the effect of SP4 orphans (duplicate Ks decays, etc.) > on the missing mass distribution? > The kinematic fitting is a very important part of your analysis. > We would like to see a comparison between data and Monte Carlo > for all variables used in the kinematic fit before the fit > For example we would like to see pmiss, costheta_pmiss, > lepton polar angle. Will plots to the BAD. > Can you please also show the pulls of the fit. > They should be centered at zero with RMS equal to one. We do not cut on P(chi^2) of the fit and make no use of the chi^2 or errors of the fitted quantities in the analysis. Therefore, the effects of underestimating resolutions in data are at most minor as far as the kinematic fit is concerned. > Has the cut value on the hadronic mass M_X < 1.6 GeV been optimised? > Please add a plot. This cut balances signal/background vs signal yield. We will vary this cut for/in the final analysis. > * Sec 4.1.1 > The starting point of the analysis seems to be a breco candidate > and a lepton with p*l > 1 GeV. > We would like to see mes plots as in Fig 1 > with a lepton with momentum p*l > 1 GeV. Added plots to BAD and added numbers to table 2. > Is the Mx binning optimal ? (probably not important for the time being) > > How is the signal efficiency epsilon^u_sel defined? Is it extracted > from an mES fit to the Vub cocktail MC or from a simple counting of > events with mES>5.27 GeV? > > What does "sideband subtraction" mean? Does it mean that the number of > signal events is taken as the integral of the signal Gaussian in the > fits to the mES distribution? Does it require a cut at 5.27 GeV in mES? > > Is there a non-(mES)-peaking component from BB-bar events in the > analysis? > > Please show a list of all cuts for the numerator and the denominator > in Equation 11. > Which cuts from table 5 are required for the denominator (b->qlv sample)? > For example do you require the kaon tag (b->u depletion)? If so, why? > > * Sec 4.1.2 > Can you please derive Eq.(12) for us? > We understand the part about the mixing, but the C and N-other term is > still not clear. Please define carefully what these are. > What is the value of C? Is it 1/2? > > * Sec 4.1.3 > It would be very useful to know the various fractions > extracted from MC. What are the values of fu, fc and foth? Also, what > are the relative fractions of the 3 components in the MC? > What are the uncertainties on f_i ? Systematic effect on these uncertainties ? > The text should mention that fc and foth are also fixed in the fit > (Is this correct?). > > We would like to have a plot of the signal shape for MX > to compare with the bkg shapes shown in Fig.35. > > * Sec 4.1.4 > Figure37 and also Figure 1, the first 2 bins: > We would expect the number of entries to be essentially constant for > the low mES bins. In all mES distributions, the first bin has > significantly fewer entries than its neighbors. > Is this due to a difference in reconstruction between the > initial processing and the final processing? > Are these bins excluded from the mes fit? The difference is due to different vertexing strategies for pre-processing and analysis processing. In Figure 1, the fit extends only down to mes > 5.21. > * Sec. 4.1.5: > A discussion of the Breco tagging efficiency eps^sl_t/eps^u_t > in the the ratio of efficiencies is missing. > Please give the individual values and the ratio. > What is the error on the ratio? > > Ratio Eff_l^sl/eff_l^u > please add the individual values for eff_l^u and Eff_l^sl > The value of 0.917 depends on the model ... Size of the effect ??? > > How quickly does the iterative procedure converge? How many iterations > do you do? > > * Sec. 4.1.6: > It's hard to understand the relevance of the multiplicity > dependence of the efficiency without seeing what effect it has on the > final answer. Did you look at that? > Please replace Fig.38 by a table with the efficiency values. > > * Sec. 4.2 > On which Monte carlo sample was this check performed? > Generic, Cocktail, Signal? > If it is not Generic: > Cocktail and signal MC have much less background than the data, > is this *really* a validation? > If it is Generic: > There is still less background in MC: higher Breco efficiency, no > continuum background. Once a lepton with p* > 1 Gev is required, the purities in data and MC are very similar. From the (new) figure 2 you can read off the numbers: all data P = 57.1 +/- 0.8 gen SP3 P = 57.2 +/- 1.4 gen SP4 P = 57.9 +/- 0.7 > Please add caption for figure 39. What is the difference between the > two plots. It's described in the text: The left plot is without the correction for the b2u tail above 1.6 GeV, the right plot includes the correction (iteration). > * Sec. 4.4 > This is a very interesting and important section that needs to be > expanded. It is comforting to see such good agreement between data and > fit for mX > 1.6 GeV. In that fit, what values of Mb_clnu and Mother did > you obtain? How do these compare with MC? Since so many of the crucial > parameters are extracted from MC, we need more tests of this sort using > data. Some of the proposed tests are: > > - Please show the mX distribution for b->u depleted sample, > in the region above 1.6 GeV. > Please show a comparison between data, and Monte Carlo. > Do data and fit agree? > What is the fraction of signal that > you expect in the region below 1.6 GeV for this sample. > If we were to look at the mX < 1.6 GeV region in this b->u depleted sample, > does this leave analysis still blind? > The above is a critical test, since we can't think of too many ways to > check the analysis before unblinding. We want to see this test as > soon as possible so that we can decide if we want to look also at > the mX < 1.6 GeV region. > > - fc and f_other value, do data and MC agree? > > - mX distribution for b->u enhanced sample with mX > 1.6 GeV in the > different superblocks. Superblock 3 has much worse purity and it > probably is not modelled as well in the cocktail MC because of > high-multiplicity modes being missing. > > - Could you please show the mX distribution for events > removed by the charge correlation cut. > How do data and MC compare? > What about fraction above and below mX of 1.6 GeV? > > - For the B+/- reco the (B+ ,lepton+) sample can be used > to check some backgrounds and the associated Mx distributions > (cascade, lepton misId) > For the B0 reco the same is possible if you solve equation 12 for > N_cascade > > - How good of a check on your efficiency could you perform on > a data sample by redoing the analysis on a B->D0lnuX sample? > In 50 fb^-1 you would have about 60 k of these events in the > D0->K-pi+. By adding additional D0 and maybe D+ decay modes > you might obtain 200 k B->D0lnuX events which could be sufficient > to check your efficiency to about 10%. > > - Since muons have more fakes than electrons we would like > to see the tests suggested above done separately for electrons and muons. > > - Any other data-based cross checks are welcome! > > Section 6 > ========= > > We would like to see a table with a list of all the systematic errors, > and their values. This has been added to the current version of the BAD. > What systematic errors are expected to dominate? In no particular order: o Fermi motion o B->X_clnu modeling o Description of PID efficiency/misid (Kaons) and low-Mx leakage o MC statistics > How are the effects of final state radiation and > Bremsstrahlung in the detector for electrons accounted for? Since we do not measure the electron spectrum, we are not very sensitive to this matter. At the momen, we use the MC simulation (SP4) to determine these effects.