Minutes of the Muon Software Meeting 11 December 2002, taken by Victoria.
 
 
0. All : Discussion of the BMU cuts
 
We had a discussion about the default cuts for high-pt BMU muons Camille
said the that the cuts that she had used for Z-analysis should be a good
starting point. Ken pointed out that many of the cuts will be the same as
for the other muon subsystems, with the exception of the dx cut and the
minimum-ionizing cuts.  Camille had been using a dx of 10cm, which
everyone agreed was probably too wide.  Camille also noted that she had
not been any significant variation of dx with respect to phi, which
indicates that the alignment of the BMU is good.
 
 
1. Volker Drollinger - chisq
 
Volker's analysis is detailed in note 6201. The aim was to construct a
chisq variables based on the the dx and dphi track-stub matching
variables.  Both j/psi and high pt muons samples were used to cover the
whole momentum range.
 
The width of the dx and dphi distributions was fitted to the function a +
b/pt.
 
Two independent variables were constructed:
chisq' = {2*sigma(dX) - sqrt(3)*sigma(dphi)}/n1
chisq''= {sigma(dphi)}/n2
Tom pointed out that this decorrelation assumes that a uniform material
density, which is not true for the CDF muon detectors.
 
The chisq variable was constructed as chisq = {sum(chisq') +sum(chisq'')},
where the sums are over all the stubs associated to the muons.
 
Volker also show us a distribution of the z0 which a second peak around
z0=1000cm.  Michael said he hadn't see anything like this!  Ken suggested
that as the value was around 1000, it might be due to a default value of
999.
 
Michael said the it would be useful to see the efficiency of the chisq cut
compared to the efficiency of the current dx and dphi cuts.
 
David Dagenhart pointed out that the new chisq cut will help to remove the
current bias we have towards selecting high-pt muons.
 
 
2. Ken Bloom, muon ID and reconstruction efficiencies
 
Ken presented some work that he will show at the joint meeting on Friday.
Using the second legs of Z events he is calculating the efficiency of the
muon selection cuts.  He also uses stubless muons to find second legs.
Compared to his last presentation on this subject he is now using Michaels
good run list.  He also warned that the cosmic tagging is not quite right
in this sample, but he doesn't believe that the ere is much cosmic
contamination is his sample. CMU wedge 17 is excluded from this analysis.
 
Using CMUP muons he gets 1478 second legs to use for efficiency studies.
There are no same-sign pairs in the 80 to 100 GeV/c^2 mass range, so this
sample is fairly clean.  Using the second legs Ken made (N-1) plots.
Most of these distributions look as expected, except for the CMU dx
distribution, which has long tails.  The overall efficiency for CMUP is
0.924. The individual cut efficiencies obtained are similar to those we
had for ICHEP, expect for the dx(CMU) cut which has improved from 0.928 to
0.965!
 
Ken also looked at the efficiencies as a function of various quantities.
In particular the efficiency decreases as the isolation and the isolation
ratio increase.  Also the efficiency seems to show some effect as a
function of run period.
 
Ken repeated the analysis using CMX muons.  There he obtained 790 second
legs to use.  The overall efficiency for the CMX legs on 0.948.
 
Ken also calculated the reconstruction efficiency using the second legs of
Z.  For this analysis he has to real muons with stubs so he can be sure
that the event would have passed the trigger. He requires that the muons
are minimum ionizing (to remove same sign contamination) and fudicial
(using a fiducial distance of 0) and then asks is there a stub in that
detector.
 
The overall efficiency is 0.954 for CMU and CMP, 0.899 for CMX and only
0.610 for BMU.  The CMP and CMP efficiencies are similar to the values
obtained for ICHEP.  The CMX is lower, but there was no attempt made to
remove runs where the CMX was off, which helps to explain the low value.
 
Michael has looked at the stability of the efficiencies.  He has not made
a rigorous calculation of the efficiencies as Ken has.  Instead Michael has
applied all but one of the cuts, which leaves him with a 95% pure sample
of muons.  Then he looks at the fraction of these rejected by the cut.
This is a good approximation of the efficiency of the cut, and it's fine
for looking at the efficiency stability.  He has only used CMP muons for
this study so far.

The efficiencies stability is checked with respect to run period (here things
look relatively stable), as a function of phi, eta and pt.
 
At high-pt it is a general feature the the efficiency at higher pt is
lower for the rest of the pt range.  Michael explained this as the
composition of the "muon" sample at higher pt has a higher proportion of
background.
 
 
3.  Ken & Victoria:  What we still need to do...
 
Ken listed the tasks that still need to be tackled before the work of the
muon software group is over.  Please see his slides for details.
 
The main discussion was on the issue of the chisq:  now G3X has been
retired, we need a reliable track-stub chisq variable.  Tom pointed out
this will be needed soon for B->mu+mu- analysis, which will start after we
have 200 fb^-1 of data.  Also B-mixing analysis will need this.  Andreas
said that a parameterization of the muon geometry in the style of SiliMap,
which would be faster than the G3X implementation.  We agreed that Volkers
parameterization of the chisq would be a useful tool for the moment.