Hi Vera, all,
> For one of our core analyses, B to D tau nu, the statistical errors
> of the MC are
> B+- B0
> PDF shape 11.5% 4.5%
> Constraints 4.2% 6.1%
> Total MC stat 12.2% 7.6%
> compared to total systematic errors of 15.6% and 9.4%.
> So, by more than doubling the MC sample, we could reduce this
> error by a factor of 1.5, in parallel to improvements in our
> understanding of the background modeling.
I won't argue with your proposal since it sounds very reasonable to me...
but I just wanted to point out that in the Dtaunu analysis, these errors
don't come from generic MC (mostly) but from signal samples. The generic
MC is in there a bit due to combinatorial BG and some of the weird
quasi-combinatorial modes, but these numbers are dominated by the signal
samples, where it's much easier to improve things.
-michael
|