Hello Daniele,
thanks very much for your information on the evaluation of Rc
from the fits.
I presume the two fits listed are the same fits as listed in Table 11,
just translating the fit results into Rc rather than Ru.  Is this what was done?

In looking at the fits, I do not understand the errors 
  a) compared to Table 11
  b) for Rc from the depleted sample and the various subsamples;
     the weighted average of the subsamples are not consistent with the
     results from the total sample and its error
  c) the "other" background is taken from MC,
     how different is C_o in the enhanced and depleted samples?
  d) when you fit the depleted sample, did you fix C_u?

What are the errors you quote?  MINUIT parabolic errors?
Have you looked at MINOS errors?  Are you constraining the coefficients
C_u, C_c, C_o to be positive?  This would introduce asymmetric errors!
Have you looked at the following information:
  - chisq vs fit parameters, 1 sigma contours, etc.
  - value and errors of the other fit parameters for various fits
All this can be extracted from standard MINUIT output options.

I am still not quite sure what is done with the first bin:
we add all data and MC up to the Mx cut into one bin, ok?
but is that bin then treated in the fit un the same way as all other bins?   

In looking at all this, I am coming back to the earlier suggestion that we should
perform a combined fit to the enhanced and depleted samples.  The depleted 
sample should fix the b --> c background, the enriched sample to extract the b--> u
signal. One would need to be careful to treat the errors and correlations correctly.
Experienced users of MINUIT could assist here.  
This should allow us to understand the correlations between C_s and C_c!
This should also help to estimate and limit the uncertainty on the s.l. branching ratios by checking the fit quality for different assumptions on the BR, not just the change in fit values.

BTW:  On page 62, Eq. 23 the parameters have incorrect subscripts,
also, it should be made clear which interval in M_x is actually fitted.
The statement about the signal events N_u=... is fuzzy as to the use
and its relation to C_s!!
Likewise, Eq. 26 and the reasoning about the other backgrounds remains very obscure!  \
If Eq is exact as you state (which it is not), then please explain
the rest more clearly.  This is best done to write the exact formulation,
and then show the simplification you chose.  I tried this in the earlier version, but
apparently I did not agree with Riccardo's view of this problem.

On a totally different subject:  Figure 6b
Shouldn't we expect to see a peaking here from this cross feed?
Valerie apparently sees a small peak leading to a correction!?

Thanks again for input,
Vera
  
 






-----Original Message-----
From: Del Re, Daniele 
Sent: Thursday, January 02, 2003 11:57 AM
To: vub-recoil
Subject: fitted BR for b->clnu events



Hi,

 first of all happy New Year.


 As requested by Vera, I produced the results for the fitted
BR for b->clnu events.

 Here you can find all numbers for the different subsamples.
 The measured values are still ratios of branching ratios, but now
the expected result is 1.

 Actually it is not equal to 1 but to 1 - ratio(BR)_bulnu (since
in the numerator we have the b->clnu component while in the
denominator we have the total semileptonic BR). Then the expected
value is ~0.98.

enriched sample
---------------

All  BRBR = 0.931841 +- 0.055113(stat)

B0   BRBR = 0.931948 +- 0.101485(stat)
Bch  BRBR = 0.949709 +- 0.0652322(stat)
ele  BRBR = 0.914472 +- 0.0692768(stat)
mu   BRBR = 0.917701 +- 0.0837971(stat)
run1 BRBR = 1.0263 +- 0.0952589(stat)
run2 BRBR = 0.890657 +- 0.0607272(stat)
sb1  BRBR = 1.01698 +- 0.103764(stat)
sb2  BRBR = 0.985293 +- 0.0750252(stat)
sb3  BRBR = 0.86291 +- 0.0714007(stat)


depleted sample
---------------

All  BRBR = 0.947103 +- 0.0280162(stat)

B0   BRBR = 0.944411 +- 0.0562612(stat)
Bch  BRBR = 0.965159 +- 0.0328286(stat)
ele  BRBR = 0.9542 +- 0.0375441(stat)
mu   BRBR = 0.923164 +- 0.0381606(stat)
run1 BRBR = 1.01548 +- 0.0475409(stat)
run2 BRBR = 0.920063 +- 0.030727(stat)
sb1  BRBR = 0.968563 +- 0.060525(stat)
sb2  BRBR = 0.998408 +- 0.0395365(stat)
sb3  BRBR = 0.909976 +- 0.0380277(stat)



 I did not calculate the error due to the MC statistics but it
should be of the same order of magnitude of the statistical one.


 These results are in agreement with the expected values for all the
subsamples. This implies that the efficiencies on the background
are well estimated.


 Daniele