The following are comments/suggestions by Eckhard an me to enhance analysis and 
clarify the documentation in BAD540:

1) check if replacing MM^2 cut by MM^2/Emiss=(Emiss-|pmiss|) is beneficial, since it takes 
    into account the resolution in MM^2. which is proportional to Emiss!

2)  check Mx fits, in particular the use of MM^2 distribution for event by event error estimate,
     what sample is used? effect of tails?  check with Oliver!

3)  add more comments on Table 8

4)  Table 9:  what is the reason for the large contributions of K+- for B+-  ??

5)  In tables 10 and 11, the numbers for BGsl for e and mu cannot be explained by mis-ID,
     what is the composition of this contribution?

6) for consistency on method, measure Rc/Rsl= BR(B--> Xc l nu)/BR(B--> X l nu) ~1,
    This ratio is know to better than 4%, and the statistics should be adequate for 5% error or better.
    Even if the answer is not right on, the variation of the subsamples, Run1 vs Run2,  e vs mu, and purity
    should tell us something! I would use the standard Mx cut and use event above
    the Mx cut in the fit.  We can then estimate the fraction of events below the Mx cut!

7) try and state the partial Ru/sl in the observed part of the phase space,
    i.e. without the correction for full Mx and p* spectrum.  This is of course complicated by
    the resolution and uncertainty in the Mx spectrum above 1.6GeV, but so be it!
    You may attempt to unfold the Mx distribution.
 
8)  Show efficiency eps^u(sel) as a function of p*, so we can correct p* distribution, if necessary

9) redo systematic error estimates:
      a) Mes fits: take into account correlations, more statistics, 
      b) check effect of Mes binning on fit
      c) check efficiency errors for e and mu (take into account time variations)
      d) the error on BR(D--> D* l nu)= 4.5 +- 0.2 % is too tight to cover the current problems,
          may be we have a BABAR measurement by Moriond!!
          BR for D(**)l nu and non-resonant D(*) l nu are correlated with this measurement!
      e) the theoretical errors are not clearly presented
            - the error on the efficiency eps^u(sel), 
                 this depends on the detector (mx fit) and decay model (angular dist, masses, lepton momentum)
            - the error on fu (fraction of events above Mx cut and below momentum cut)
            both depend on Fermi motion, i.e. on values for parameters mb and a,
            but there is also an experimental contribution to this error.  should be factorized!
      f) where does the MC statistical error come from?  the tag efficiency ratio is taken from
         MC, it error is purely statistical?

10)  The variation of Ru/sl with Mx for B0 remains a worry (recent posting by Urs needs
     correction to project uncorrelated errors correctly),
     Can we investigate what leads to this rapid change near 1.6GeV??  check data/MC !

Best regards and congratulations for a job very well done!
Merry Christmas,

Vera and Eckhard

           
         


      

============================================
Vera G. Luth                         Phone:  650 926 2702   
SLAC MS 95                         Fax:      650 926 2657
P.O. Box 20450
Stanford, CA 94309                e-mail: [log in to unmask]
============================================