Hoi,

the trk  killing was implemented with the  Daniele's tables, contained
in IslBrecoilUser/tablechg.dat, giving the ratio data/MC.

In momentum intervals with an  'efficiency' larger than 1, tracks were
killed in data.  In the  following, the usual plots are inverted: what
is labelled  as data is actually  generic MC, and  the histograms show
the killed and default data distributions.

The track momentum spectrum is well reproduced by this procedure on
the lepton-cut only sample: 
  http://www.slac.stanford.edu/~ursl/talks/070202/trkkill-a1600.eps.gz

There are  residual differences (and  fluctuations in data)  after all
cuts:
  http://www.slac.stanford.edu/~ursl/talks/070202/trkkill-a1610.eps.gz

This is a  bit surprising, I'd have thought that  the tables are after
all cuts.

The  effects of trk  killing are  clearly visible  in the  (fitted and
reco)  MX  distribution,  which  is systematically  shifted  to  lower
values:
  http://www.slac.stanford.edu/~ursl/talks/070202/trkkill-a2420.eps.gz

The agreement between data and MC is better for the unkilled version. 

The charged Trk multiplicity is affected, especially after the lepton
cut: 
  http://www.slac.stanford.edu/~ursl/talks/070202/trkkill-a4300.eps.gz
  http://www.slac.stanford.edu/~ursl/talks/070202/trkkill-a4320.eps.gz

The default setup shows better agreement. 

Note: These plots are a bit different from the BAD, this is due to the
fact that the BAD plots were made with the veryTight muon selector and
these here with tight.

So, technically, I think there no obvious problem with the TrkKilling.

The fits give: 

  Default: BRBR = 0.0167335 +- 0.00446966(stat) +- 0.00127934(MC stat)
  TrkKill: BRBR = 0.0218932 +- 0.00459749(stat) +- 0.00124435(MC stat)

This is  a difference of 31%.  Given the level of  disagreement in the
above distributions, I  would not say that it's a  good estimate for a
systematic. More thinking is required. 


Cheers,
--U.