Skip to content

Latest commit

 

History

History
70 lines (51 loc) · 2.64 KB

oscnext-nsi.md

File metadata and controls

70 lines (51 loc) · 2.64 KB
Raw

OscNext NSI

An ongoing IceCube NSI analysis (wiki).

Follows the precedent of the DRAGON NSI analysis.

Performed by Elisa Lohfink (see Elisa-thesis-2023).

Uses 8 years of DeepCore data from 5.6 to 100 GeV (Elisa-thesis-2023).

[[PISA]] on its own cannot handle the NSI parameter space for a few reasons. MCMC is used in place of minimization via emcee. This will allow simultaneous evaluation of all NSI parameters (Elisa-thesis-2023).

Uses $$ TS = \chi^2_\text{mod} = \sum_{i \in \text{bins}}\frac{(n_i^\text{exp} - n_i^\text{obs})^2}{n_i^\text{exp} + (\sigma_i^\text{exp})^2} + \sum_{j \in \text{prior}}\frac{(\Delta s_j)^2}{\sigma_{s_j}^2} $$ (Elisa-thesis-2023)

Status of the analysis when I received it

Frequentist:

  • Blind fit p-values:
Parameter p-value
$\epsilon_{e\mu}^\oplus$ 4.4%
$\epsilon_{e\tau}^\oplus$ 4.6%
$\epsilon_{\mu\tau}^\oplus$ 5.4%
$\epsilon_{ee}^\oplus - \epsilon_{\mu\mu}^\oplus$ not fit
$\epsilon_{\tau\tau}^\oplus - \epsilon_{\mu\mu}^\oplus$ 4.8%

Bayesian:

  • Performed MCMC with all magnitudes free, but not phases, on highstats

Steps forward

Frequentist:

  • Switch to using dedicated systematic sets for each NSI parameters
  • Test minimizer configuration
    • Shiqi: talk to astronomer postdocs on the third floor
  • Improve FC computation
    • See work by Elisa and Alex
  • Include $\epsilon_{ee}^\oplus - \epsilon_{\mu\mu}^\oplus$? (see Elisa 7.6.2) (currently excluded for computational reasons)

Bayesian:

  • Improve computational efficiency, if possible (currently at 800 CPU hours per sampling run)
  • Improve sensitivity to complex phases, if possible
  • Test ultrasurfaces
  • Determine relevant checks on result
    • Shiqi suggests inject/recover, pre-fit data/MC, sensitivity
  • Perform with current nuisance parameters

General:

  • Point to FLERCNN sample
  • Extend energy range if feasible (up to 300 GeV?)
  • Switch to SPice-BFRv2 (see Elisa 7.4.2)
    • from SPice 3.2.1
    • requires new MC sets (do these exist yet?)
    • Would obsolete the $N_\text{brf}$ parameter
  • Reevaluate energy binning
  • Check whether or not cutting at the horizon makes a difference for NSI
  • Check flux binning (low priority)
    • Shiqi: after doing this, try doubling true flux and oscillations binning and check sensitivity (need to figure out what this means)
  • Check PID binning (low priority)
  • Make sure ultrasurfaces are up to date