meeting 2025 01 15 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

Context

  • Just wanted to include a few diagnostics here when we switched from all bins --> removing the 5 inner bins, rather than include these plots in emails.
  • As a quick reminder, the inner 5 GMOS bins had elevated sigma and h4, likely due to the central region's nuclear dust. I've run two dynesty tests, one which includes the chi2 from all bins, and the other which removes these 5 bins' contribution to the overall chi2.

    • The results look mostly nice, with two details worth mentioning:

      • The M/L preference drops to just outside the original 1sigma region, from 2.62+/-0.04 to 2.53+/-0.04.
      • The posterior for T has a bit more posterior mass near T=0.92 (our original minimum), and thus the shape contours are a bit strange looking in our final cornerplots

    • I tried doing a similar trick to what we had been doing for N57 a few months ago to address the difference in the recovered shapes. Instead of assuming a uniform prior on T, I instead am using a Gaussian prior with mean = 0.97 and a standard deviation = 0.02 (twice as big as the errors on T).

      • The result from this is very ncie looking, grealty reducing the posterior mass at T=0.92 to approxiamately the same level as our old "fiducial results" which included all the bins.

Plots

  • Here are some diagnostics related to the discussion above. First, here's our old "fiducial" results which had all the bins included:
T's Final Cornerplot
[images/250115/nIter8_best_grid_D_and_E_grid_alpha_K100_nu1.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250115/nIter8_best_grid_D_and_E_grid_alpha_K100_nu1.5_testing_testing_testing-1.png)
  • And here's the results with uniform priors on all parameters, but without the inner 5 bins:
T's Final Cornerplot
[images/250115/240110_inner_5_removed_no_prior_grid_alpha_K100_nu1.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250115/240110_inner_5_removed_no_prior_grid_alpha_K100_nu1.5_testing_testing_testing-1.png)
  • And lastly, here's the results applying a Gaussian prior to T, with mean 0.97 and std = 0.02.
T's Final Cornerplot
[images/250115/240110_inner_5_removed_grid_alpha_K100_nu1.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250115/240110_inner_5_removed_grid_alpha_K100_nu1.5_final-1.png)