|
Stephanie's Homepage
|
Level 6
Level 6 cuts are done in ROOT, and are run by me on local computers at Canterbury.
Cut variables
Level 6 is the final level of cuts, they are as follows:
- BDT response: MVA_BDT > 0.2
- Energy: log10(CredoFit4_Energy) > 4.4 (25.1 TeV)
Development of cuts
Figures 1 and 2 show the BDT and Energy cuts.
| 
Figure 1: BDT response. The cut is shown at MVA_BDT > 0.2 by the black vertical line. Figure 2: Reconstructed energy using 4 iteration Credo. The cut is shown at log10(CredoFit4_Energy) > 4.4 by the black vertical line.
|
Figure 3 shows the two dimentional parameter space that these cuts are optimised in. Data is used for the optimisation rather than CORSIKA since there is a clear lack of statistics in the signal region. Assuming there is some signal in the data the optimisation must be selective. Events that lie in the extreme tails of the distributions (high BDT response and high energy) are removed for the optimisation. This risks cutting out some signal events due to overly cautious optimisation of the cuts, however the method is far more robust than relying on the limited CORSIKA simulation.
Feldman-Cousins Model Rejection Factor (MRF) and Model Discovery Factor (MDF) give optimisation values of:
MRF = 0.0299752 at MVA_BDT = 0.2 log10(CredoFit4_Energy) = 4.4
MDF= 0.0282623 at MVA_BDT = 0.2 log10(CredoFit4_Energy) = 4.4
Figures 4 and 5 show this optimisation.
| 
| 
Figure 3: MVA_BDT vs. log10(CredoFit4_Energy): a) Experimental data, b) CORSIKA, c) Atmospheric Signal, d) E-2 Signal. Figure 4: Optimisation using MRF. Figure 5: Optimisation using MDF.
|
Since there are no corsika events left after final level, the background is estimated in the signal region for the full data set of IC40. This is done using the burn sample at level 5. The final cut variables at level 6 are BDTresponse and Energy. This plot of BDTresponse vs. Energy is shown in Figure 6a, the cut values of BDTresponse=0.2 and log10(Energy)=4.4 are shown by the black lines. The BDTresponse vs. Energy in the background region (lower left quadrant) is fit by a 2d function, which is then extrapolated to the signal region (upper right quadrant). This fit function is shown in Figure 6b. The quality of the fit is shown in Figire 6c, where the relative deviation of the data from the fit function is plotted. As has been shown before using 1d plots, CORSIKA and data agree in the background region well: this is illustrated in Figure 6d where (CORSIKA-fit)/CORSIKA_error is plotted.
Figure 6: Background estimate a) BDTresponse vs. Energy at level 5, b) Extrapolation, c) Deviation of fit from data, d) Deviation of fit from CORSIKA.
|
This extrapolation yields a background estimation of 0.0517351 events with an 90% upper limit of 0.345 events for the entire IC40 livetime. The upper limit is obtained from a fit to the integral valus obtained by varying the fit parameters within their uncertainties. This is hence the most conservative estimate of the background upper limit.
The passing rates for level 6 are shown in Table 1.
| . | Trigger Rate (Hz) | Level 2 Rate (Hz) | Level 3 Rate (Hz) | Level 4 Rate (Hz) | Level 5 Rate (Hz) | Final Rate (Hz) |
| Experimental data | 1500 | 16.3 (1.1%) | 1.75 (10.7%) | 2.54 × 10-2 (1.5%) | 2.09 × 10-3 (8.21%) | 6.37 × 10-7 (0.03%) |
| Monte Carlo | 1270 | 12.5 (1.0%) | 0.92 (7.4%) | 3.30 × 10-2 (3.6%) | 2.49 × 10-3 (7.54%) | 0 (0%) |
| E-2 signal | 2.55 × 10-4 | 1.48 × 10-4 (58.0%) | 1.15 × 10-4 (77.9%) | 5.55 × 10-5 (48.2%) | 1.83 × 10-5 (32.97%) | 7.79 × 10-6 (42.57%) |
Table 1: Final passing rates.
|