"If the efficiency varied with pT or y in not-understood ways
this could affect the measurement"
|pT resolution ||
"if the pT resolution was wrong in our simulations
this might affect the t spectrum"
|Detector Simulation ||
"We have studied a variety of relevant variables, and found that the
vertex position, rapidity distribution, M&pi&pi distribution, and &pi ±"
angular distributions agree well between the data and simulation. In addition, we have turned off the
detector simulation, by fitting the uncorrected t spectrum with the raw Monte Carlo output; this
reduced c by 0.18, a relatively small change...We have confidence that out detector simulation
is at least 75% correct, so we assign an overall 5% systematic error to the detector simulation."
"Backgrounds were estimated using like-sign pairs...The like-sign background
percentages with t < 0.01 GeV2...are ~ 2% of the signal. The like-sign backgrounds should
be within a factor of 2 of the true background."
F. Meissner asseses this at < 1%
according to the ZDC paper and hadron yield paper.
V. Morozov "compared the ratio of Nhits to the number of possible hits on the track
Npossible hits in data and MC. The number of possible hits was calculated by extrapolating a track's
geometry over the pad plane. The calculation incorporated both the detector geometry and features of the track finding
routine....The quality of the track fit in data and MC is characterized by the &chi2 per d.o.f of the track
fit...In a related analysis the authors studied the variation in the tracking efficiency-corrected pT
and &eta spectra due to small variations in track quality cuts. The systematic error was found to be 6.4% for each
reconstructed track, which we accept as a tracking efficiency systematic error for the present analysis." Still looking up
the reference. I don't know if it applicable in this analysis or if an analogous study exists.
has a quote of 4% per track which he explains by:
hadron yield paper: 70-95% (p_T= 0.1-2GeV), with 85% for p_T>0.2
compares well to values obtained from the rho Monte Carlo, see
|Vertex Finding ||
V. Morozov says he plots "invariant mass spectra for identified secondary
e+e- pairs comprised of any two global tracks and for pairs comprised of two global
tracks that also were primary tracks...The ratio of the e+e- pairs with primary
tracks to the total number of pairs gives an estimate of the vertex finding efficiency in the data."
He comes up with an 8.5% estimate of the error in vertex finding efficiency.
has a quote of 8% which he explains by:
Compare LMV and DCA for the rho MC, see
|Pair Identification ||.
V. Morozov estimates the "systematic uncertainty in pair identification efficiency
via specific energy loss". He goes through a rigorous mathematical justification to come up with an estimate of 2%
which ends up being negligible in comparison to the statistical error.
|pT Mis-Reconstruction ||
Morozov says in order "to study the systematic
error due to shift in reconstructed pTreco compared to the true value of transverse
momenta," he re-computes "the reconstruction effficienty without applying the shift correction to the
simulations. The raw Minv distribution (which is computed with pT correction)
is then efficiency-corrrected with two different efficiency functions - one with account of pT
shift, and the other without." Subsequently, using the error of the pT correction to be
~10% he estimates the error due to the pT distortion to be ~2%.
To estimate whether the number of reconstructed tracks in the East vs.
West side of the TPC is different Morozov computes the ratio:
A = (N++ - N--)/(N++ + N--) x (N++ + N--)/(N++ + N-- + N+-)
where N++ is the number of events in the identified pool of events with two tracks with
positive pz . The ratio ends up being within 2&sigma of zero, and therefore not appreciable.
This may be akin to my own zVertex studies.
|Sensitivity to Generated Spectra ||
Morozov attempts to account for the possibility that the physics model
of the event generator is wrong. "To test the sensitivity of reconstruction efficienty to variations in
generated spectra, we can generate an event sample with, for instance, flat pTtot pass this event sample
through the GEANT simulation, event reconstruction and analysis, and determine reconstruction efficiency in that way. Alternatively,
we can use the same event generator event sample that we have been using so far(with pair transverse momentum distributed
according to the probability distribution f(pTtot)), but compute the efficiency by the formula:
Reconstruction Efficiency From Flat pTtot Spectrum =
= # weighted events after detector simulation and all analysis cuts /
# weighted generated events in acceptance"
where the weight for each event is inversely proportional to f(pTtot)). He ultimately determines that
the reconstruction efficiency "is not affected by the shape of the generated distribution" for most values of the transverse
|Varying All Analysis Cuts ||
Morozov varied all cuts by ± 2.5% ± 5.0% ± 10%. He ultimately
determined that "the relative numer variations were found to be very close (differ on the average by ~ 4.5%) in data
and simulations, therefore the simulations describe the data response to the analysis cuts with excellent accuracy".
Meissner did a cut variation study to which he attributed a 5% systematic error.
Supporting plots here.