Hello folks,
first of all I realized that we never estimated the significance of this
|Q|<=1 change. The difference in quadrature of the errors does not
work if both signal and background go up. I then realized we never lloked
at the result for just |Q|=1 events.
The result is
BRBR = 0.057 +- 0.017
to be compared with the completely independent default result
BRBR=0.0170+- 0.0043

The two measurements are 2.2 sigma's apart, which is the size of the
effect we are discussing.

Moreover the fit result with |Q|=1
http://www.slac.stanford.edu/~rfaccini/phys/vub/q1fitresults_nocat.eps
shows that the fitted distribution does not fall off too well below 1.6
GeV (although the chi^2 of the fit is good because we lump Mx<1.55
altogether), most of the signal beeing clustered at 1.5GeV.

I also performed a test on the toy MC, as already previously described.
I fitted (101 means |Q|<=1, qhile q1 means |Q|=1)
q0   http://www.slac.stanford.edu/~rfaccini/phys/vub/q0toy.eps
q101 http://www.slac.stanford.edu/~rfaccini/phys/vub/q101toy.eps
q1   http://www.slac.stanford.edu/~rfaccini/phys/vub/q1.eps

Resolutions in q0 and q101 are the same as in data demonstrating the the
toy is realistic.

The distribution of Q1 alone is far from being gaussian and actually
the probability of getting the value we get (0.056) is quite high [you
must also consider that the toy has a generation value which is 0.05 lower
than what we measure on data. In the plot there are 5% of the trials above
the one corresponding to data.

As a check I plotted the error on BRBR in the q1 sample
http://www.slac.stanford.edu/~rfaccini/phys/vub/sigmaq1toy.eps
which compares well with the error on data (0.017).

 Overall the effect seems to have a ~5% probability
	Ric