TWIST Analysis Plans
Last Updated October 16, 2003
See also the TWIST
Systematics Studies.
- Helix Fitter Tests
- Tuning of drift time/distance fitting code
(Konstantin/Andrei). This will be done after blind analysis
code is prepared. Present helix fitter vs michel parameter
sensitivity:
- fit with L/R unknown
- fit with L/R from MC
- compare b & c
- fit with wire centers -done
- compare e & c/b
- fit with different strs
- move/stagger chambers in GEANT -done
- with and without fieldmap
- Pattern Recognition Tests (Jim/Blair)
- Study effect of muon dead zone on classification.
- GEANT Test - Artificially deaden hits some number of
planes before muon stop. See at what level this effects
michel parameters compared to non-deadened run.
- Data test - Compare classification of early
non-overlapping decays (1000 - 1500 ns) to late decays
(5000 - 9000 ns).
- Study inaccurate winding number assignments.
- Apply early rough cut, such as average wires/plane or cluster
size/pair. Examine effect of throwing out outliers.
Amost done see firstguess postings.
- Use cluster size vs. |cos(theta)| relationship to aid
in FirstGuess.
- Use to constrain winding number in (omega_z, phi_not) fit in
manner similar to dense stack constraints.
- Evaluate resulting cluster size vs. |cos(theta)| for accuracy of
fit.
- Evaluate effect of instrumenting more of the dense stack (Jim).
- Study inefficiencies in delta removal code.
- Compare reconstruction of GEANT runs with and without delta production.
Eventually should be full statistics study to see sensitivity to
Michel parameters.
- Explore possible correlations with delta removal failures.
Success/failure vs. (MC decay params, delta parameters, other)
- GEANT-to-GEANT Studies
- Evaluate backgrounds from showering in the yoke - not possible
using field map.
- Study October 23, 2002 beam tune change by fitting set B to set A.
- Use set A as standard distributions.
- Use GEANT produced derivative distributions.
- Test classification for case if target PCs are used in ADC mode, and so
have low efficiency for TDC information. (Do test by throwing away all
target PC TDC information.) Progress...
Done. No effect on Michel in low statistics test.
- Examine effect of changing parameters in classification. Check
sensitivities for both GEANT and data.
- Cross Talk, Noise, and Cosmic Removal Parameters
- Windowing Parameters
- Other Side Delta Removal Parameters
- High Angle Track and Scattering Detection Parameters
- Muon ID Parameters
- Soft Uncorrelated Laserlike Track ID Parameters
- Ntuple updates... not pattern recognition(Blair)
- Debug ID of pi-> e nu windows to make sure they make it to the ntuple
Done? (21 Aug 2003)
- Post list of ntuple variables, get feedback, and add/remove variables
Done, here's the list (7 Jul 2003)
- And some root tree documentation Documented Here.
- Asymmetry/Muon Polarization Studies
- Check sensitivity of asymmetry estimate to different
cuts Done with no clear culprit. Was tune
even worse than we thought?
- Try other methods of analysis developed
- Look at asymmetry of different data sets
- Slightly upstream and slightly downstream
- Changed B2
- Different densities
- Different stopping locations (different C02%
or extra foils)
- Revisit T1GEANT to try to estimate expected muon
polarization for stainless steel cladded Be target.
- Get some new data with Graphite target
- Get data to study polarization as muon momentum is lowered
(different depths in production target).
- Blind Analysis Code Preparation (Andrei)
- Develop Fitting scheme Done. Doc required.
- Ripple studies
- Check for effect of different binning - Done, biases remain
- Artificially worsen resolution
- Use alignment files for generation of GEANT - no conclusion
whether to make this the default.
- TWIST Geant Tests (Rob et al.)
Ways to verify Geant against real data. The main person looking into
each item is listed in [square brackes]. Current priorities are listed for
RPM's tasks.
High Level Tests
These are looking at various impacts on the "end results" of
simulations, especially momentum and angle distributions. Michel
spectrum fitting will be useful for some of these.
- "Downstream Aluminum Problem" [RPM (1)]
- Understanding why an aluminum plate placed downstream has
so much more impact in data than in Geant.
- "Geant vs Geant" [RPM (1.1)]
- Mirrors of various "data vs data" studies.
- Still looking at 'downstream aluminum'
studies. Investigate delta removal.
- Coordinating with JRM to ID studies that show the
largest effects (DS aluminum, DC HV, etc).
- A list of relavent forum
links is available.
- Energy Loss & Multiple Scattering [RPM
(3)]
- Can also compare both Geant and data against theoretical
calculations. Especially when all material thicknesses are
known.
- I'll use muon ranging during momentum scans to get
another estimate of the target's graphite layer
thickness, to compare with that suggested by the energy
loss results.
- The aluminum target (with known thickness) will be very
valuable for these tests, as well.
- "Slightly" upstream/downstream stops.
[RPM (5)]
- Test response (e.g. asymmetry, momentum distribution, etc.) to
muons stopping in target, but not centred.
- Energy calibration with pions. [Roman]
- Compare measured response function of 70 MeV/c positrons from
decay of stopped pions.
Low Level Tests
These are looking at parts of the simulation in detail.
- Distribution of cluster sizes and numbers
of hits. [RPM (2)] (Clear evidence of extra
hits not modeled in GEANT, UV photons still a plausible
mechanism.)
- Mainly a test of clustering, and of the efficiency of detecting
when a track clips a cell corner.
- Cluster size discrepancies can skew some parts of the
reconstruction, so it's important to get them
right (or at least close).
- A list of
relavent forum links is available.
- Reconstruction Efficiency [RPM (4)]
- Looking at through-going tracks (e.g. decays from
upstream-stopping muons); how many reconstruct in one half but
not the other? Compare distributions between Geant and data.
- TDC spectra. [KO?]
- Test of clustering, resolution functions, etc.
Back to TWIST Software Page
To the TWIST
Geant page.