meeting 2024 10 08 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

Context

  • I've reminimized a host of scalings (+/-0.5%, +/-1%, +/-2%, +/-3%) for Grid A (866 models) after having masked the outer 4 Mitchell bins and the innermost Mitchell bin which overlapped with the GMOS data. Again as a quick reminder, these models have the outer binning issue fixed, and were minimized using the newest sets of triaxnnls binaries.
  • The results are looking very similar to what we've seen before. At this stage, Grid A doesn't provide much coverage and the scalings don't drastically alter what we're seeing. This is about the same as we saw when we first started working on N315, with the majority of well-fititng points coming from Grids C/D.
  • Other interesting takeaways:
    • The NNLS chi2 are noticably "smoother" between adjacent scalings. This isn't that surprising, but it definitely stood out to me. I think this is just a further sign to stick with the NNLS chi2.
    • I think we're too undersampled at this point to do anything meaningful with GPR and dynesty so I've held off for now.

Plots

  • Here's a quick comparison of the 1d panels in T space vs. angle space:
Case T Space Angles Plot from 1/30
NNLS
Kinem
  • And here are the 1d panels for the scalings individually:
scale=0.97 scale=0.98 scale=0.99 scale=0.995 scale=1.0 scale=1.005 scale=1.01 scale=1.02 scale=1.03
NNLS
Kinem
  • A quick comparison between the kinem chi2 and the NNLS chi2 indicates that these are tracking each other quite well:
Plot
  • And just for reference, here are the 1d panels considering only the best selected scales, and the corresponding distribution of scales which get selected. It looks like the data preferentially prefer higher and lower scalings? But again, we're greatly limited in our coverage so unsure that we need to read that deeply into this plot with how things currently stand:
Panels Distribution

Plan Moving Forward

  • I wanted to quickly remind ourselves of what we did after the Grid A exploration with the original N315 minimization so that we can decide what to do from here.

    • After wrapping up the scalings from Grid A, we had submitted a grid of d1 models over the range:

      • BH = [5e8, 4e9]
      • ML = [2.0, 3.2]
      • rho = [1.7e9 6.5e9]
      • T = [0.5,0.99]
      • Tmaj = [0.01,0.5]
      • Tmin = [0.4,0.99]

    • We then selected the models which were within 1000 of the d1 minimum, leaving us with 729 new models to run with d3. In the results below, I've also added 6 scalings to this grid, at 1, 3, and 5 percent above and below the base set of models.

  • I think we're fine to continue on and re-minimize the kinematics we are currently using with these same sets of models. It might not be the most ideal case (since this was suited for the old set of kinematics), but the re-minimizations are cheap and this will at the very least fill in the space with about double the models. I think going from Grid B --> Grid C is where we could start to see some difference, since Grid C was not uniformed (it was instead based on rejection sampling).

Here are the 1d panels showing the results of those tests (from the 02/06 page):
729 Base Models All Scalings (729*7) Best Scalings(729 Models)

Other Diagnostics

Here's the current best-fitting model diagnostics + kinematics with one very strange thing that I discovered this AM and am still trying to diagnose -- for some reason TriOS is reading h3-h8 for the innermost bin in some bizarre way from the kindata.dat file, and the inputs for h3-h8 are being output as 0 with large uncertainties. The innermost bin h3-h8 is the only spot where this is the case, and I'm still trying to figure out where this is coming from since the kindata.dat files look perfectly fine. Because the errors are so large, this shouldn't impact our results at all but just wanted to keep this in mind when viewing these plots.
Kinematics Beta + Orbits
⚠️ **GitHub.com Fallback** ⚠️