General numu analysis details¶
0) Useful Links:
Most of this information can be found in other places, probably in greater detail. Some important resources:
executive summary of FA for box opening: http://nova-docdb.fnal.gov:8080/cgi-bin/ShowDocument?docid=13641
old tech note of FA http://nova-docdb.fnal.gov:8080/cgi-bin/ShowDocument?docid=12118
analysis summary for SA: http://nova-docdb.fnal.gov:8080/cgi-bin/ShowDocument?docid=15232
The numu analysis looks for the deficit of muon neutrinos at the FD. The basic analysis design is to find neutrino events by removing backgrounds (primarily cosmic ray muons at the FD, and neutral current events at both the ND and FD), and only look at events that are fully contained inside the detector(s) so we can most accurately reconstruct their energy information. We then look at simulated MC at different oscillation assumptions (extrapolated from the ND) and see which oscillation assumptions best match the data with a spectral chi-sq fit.
The meat of the analysis work is in designing the cuts to remove enough background yet maintain signal efficiency; to find ways to better reconstruct the energy; to find ways to lessen the impact of systematic uncertainties; to ensure all uncertainties are adequately covered; and to validate technical implementation of all details.
The first analysis suffered from the hadronic energy problem and had to 'correct' hadronic energy with a 14% shift (applied to data). Second analysis used simulation with Dytman MEC events and did not need such a shift, and had updated systematics. Third analysis will revisit all aspects of the analysis and potentially improve them.
The cuts developed for FA were not changed for SA. They will all be revisited for third analysis. They currently are:
(using the same standard spill cuts as with nue)
a) Quality cuts
- this basically requires at least 20 hits in the slice and that we have some Kalman tracks with real ReMId values. This cut is the same in the ND and the FD.
- this ensures our events are fully contained so we can accurately reconstruct the energy. It also removes almost 99% of cosmics in the FD. The containment has a part based on tracks, looking at the projections along the track direction to the edge of the detector, and a part for hadronics, which looks at how close any hit in the slice is to the edge. Though the philosophy is the same for ND and FD, the actual cuts are different, and the ND has an extra part dealing with the muon catcher.
c) NC rejection
- NC events are rejected using the ReMId PID, a 4 parameter kNN. The ReMId cut is the same in both the ND and the FD, and succeeds in removing most NC events. This cut has the largest impact on signal efficiency, especially at low energy.
d) Cosmic rejection
- To start with we need about 7 orders of magnitude reduction is cosmics to do the analysis. Even after all other cuts we still need almost 3 orders of magnitude. This is accomplished with a BDT, applied to the FD only. Different BDTs are tuned for the period 1 data and for period 2. Period 3 used the period 2 tuning. The cut values are the same even for the different tunings.
The overall cut efficiency, 0-5 GeV, for truly contained CC events, is around 60%. Almost all our sensitivity is from < 2.5 GeV.