[tomaskontrimas]

Fixes the flux conversion in the likelihood scan tutorial (doc/sphinx/tutorials/likelihood_scan.ipynb) reported in #336, plus a few related corrections.

Flux conversion (#336)

The tutorial computed the flux model scaling factor once at the best-fit spectral index $\hat\gamma$ and reused it for the whole $(\Phi_0, \gamma)$ grid. Since the factor depends on $\gamma$, every scan point with $\gamma \neq \hat\gamma$ was placed at the wrong flux normalization, and the contours were only correct at $\hat\gamma$. The factor is now calculated for each $\gamma$ of the grid.

Other fixes

• Evaluate the LLH ratio function at the fitted parameter values instead of the hard-coded [80.09, 3.21], and show that $\mathrm{TS} = 2 \times$ llhratio_value.
• Mark the best fit at the fitted values (using mu2flux) instead of the grid minimum, and use the unblinded TS as the best-fit TS.
• Use the geometric mean for the centers of the log-spaced flux bins.
• Correct the $\Delta\log\mathcal{L}$ equation, which was missing the factor of 2, and follow the $-2\Delta\log\mathcal{L}$ convention of the NGC 1068 paper (arXiv:2211.09972) in the text and the colorbar label.
• Make the best-fit marker visible in the legend, remove leftover _flux suffixes from variable names, and fix typos.

[Copilot]

Copilot wasn't able to review any files in this pull request. Check if the Files changed in this pull request are included in default exclusions.