Hoi Ric,

 > - what happens if you require the already suggested cut Ndch>0 if pt<200
 > MeV? Does the effect disappear? I guess slide 3 should tell me something
 > but I do not know what you are plotting.

You mean loopers? No. Most of the loopers are SVT-only pieces not
linked to the main track. If a track exits the tracking volume, it
will start a new track.

> slide 3

page 2? Delta phi as calculated between close candidates in delta pT
and delta theta. 'Close' means 200 MeV for pT and 0.2 rad for theta
(and v.v. 0.3 in delta phi when displaying delta theta). These are
just loose cuts for displaying, the matching cuts for looper searches
are tighter (and at the moment 'by eye'). In the left most column you
see conversion (the peak at delta phi = zero) and loopers (the peaks
at delta phi = +/-pi). For the theta plots, only the loopers are
shown. All the plots on page 2 (as indicated at the bottom of the
page, not in acroread) are on data.

Ghost tracks are tracks with pT > 200 MeV, same sign, same phi, same
theta. Same distributions as for loopers on preceding page. I don't
yet understand the phi distribution in data. It looks different from
MC.

 > - what do you mean efficiency is low? I cannot understand which cut you
 > are suggesting to apply and what is its efficiency.

I don't say it's low. All I say is that there is a resiudal bump in
data at theta = 90 degree which is more significant in data than in
MC.

 > - I do not understand how does the material you show quantify the issue
 > on Qtot. Which is the fraction of events where the loopers have an impact on
 > the charge reconstruction? How well is it reproduced between data and MC?
 > Applying an ad hoc selection how well can we do in removing the problem?

Remember, it's not a final result ... Loopers have an impact on Qtot
whenever the 'No. of Duplicates' on page 6 is odd. The total number of
events processed in data was 596330, the plot (top left) seems to
indicate something like 25-30k tracks where one remaining os-looper is
killed.

Note that in the distributions I have not disentangled BRECO and
recoil. Loopers are also contained in the BRECO tracks (cannot
quantify at the moment). This might be an entirely different story.

I admit that I have a conceptual problem with the combination of the
Qtot cut and the tracking features alluded to in my posting. The gain
of having the statistical error on Vub go from (say) 10% to 9% (the
10% improvement is based on a pure MC study) but having to prove to
ourselves satisfactorily that the MC does a good (enough) job
describing problems (not features) in tracking seems marginal.


Cheers,
--U.