meeting 2024 01 30 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

Context

• I ran the full set of scalings that we were discusing over email, and here are some of those diagnostics.

• As a quick reminder, we've been dealing with 866 base models over the range:

  • BH = [5e8, 5e9]
  • ML = [1,4]
  • rho0 = [1e9, 4e9] [in code units which is a factor of 4*pi larger than physical units]
  • T = [0.01,0.99]
  • Tmaj = [0.01,0.99]
  • Tmin = [0.01,0.99]

• I've added 10 scalings, giving us 11*866 = 9526 models total. These scalings are:

  • +/- 1%
  • +/- 2%
  • +/- 3%
  • +/- 5%
  • +/- 8%

Diagnostics

• Considering all the points together, here's a breakdown of the number of models as a function of cutoff K:

Cutoff K	Num. Models < min(chi2) + K
10	3
50	10
100	24
250	269
500	1657
750	3633
1000	5421
1500	7822
2000	8827
5000	9526

• Considering only the best scaled version:

Cutoff K	Num. Models < min(chi2) + K
10	2
50	4
100	10
250	81
500	365
750	621
1000	762
1500	858
2000	866

1d landscapes

• Here's the 1d landscapes for all of the model points vs. only the best version of each scaling

Best Scales	All Scales
[images/240130/1d_panels_best_scalings.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/240130/1d_panels_eleven_scalings.png)

Dynesty Posteriors

• I've also run a set of gpr/dynesty runs on the set of best-scaled models:

Cutoff K	nu=0.5	nu=1.5
500 (365 model points)	[images/240130/240130_d3_best_scalings_grid_alpha_K500_nu0.5-1.png]] ](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/240130/240130_d3_best_scalings_grid_alpha_K500_nu1.5-1.png)
750 (621 model points)	[images/240130/240130_d3_best_scalings_grid_alpha_K750_nu0.5-1.png]] ](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/240130/240130_d3_best_scalings_grid_alpha_K750_nu1.5-1.png)
1000(762 model points)	[images/240130/240130_d3_best_scalings_grid_alpha_K1000_nu0.5-1.png]] ](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/240130/240130_d3_best_scalings_grid_alpha_K1000_nu1.5-1.png)