 |
 |
Hi all, please find at http://www.slac.stanford.edu/~bozzi/MCtests/MCtest_bchlpt.html the results of further studies on MC based on the 3 fitting algorithms we have been using and discussing so far, and on some subsamples (charged/neutral B, electrons/muons). The summary is that the method where we apply a flat correction for the peaking backgrond is badly biased and prone to fluctuations. The one where bin-by-bin corrections (in mx or P+) are applied is bias-free and very stable. The approach where all (data and MC) mES distributions are fit is almost unbiased although there are some fluctuations here and there. We propose that we drop the approach with the flat correction and concentrate on the other two. The "fit all" approach has a bigger MC stats error than the "correct all" approach, but the latter suffers from the systematics due to the peaking background knowledge. The overall uncertainties are comparable. We will now start to switch on the various corrections, and use both approaches. One of them will be our default, the other will be used as a cross-check. Cheers, Antonio and Concezio.