meeting 2025 03 12 n315 - JacobPilawa/TriaxSchwarzschild_wiki

Context

We ran three additional scalings on Grid E at the end of last week, and here are the results:
- Just as a quick reminder, this brings the model coverage to (0.985, 0.99, 0.995, 1.0, 1.005, 1.01, 1.015) for Grid D, and (0.985, 0.99, 0.995, 1.0, 1.005, 1.01, 1.015, 1.02, 1.025, 1.03) for Grid E. The kinematics here are the updated + properly masked kinematics.
Takeaways/Results:
- The additional scalings didn't fully resolve the coverage of Tmaj/Tmin's 3sigma region, but again I'm not sure we should be too concerned at this point since we're mostly focusing on the angles and shapes for this paper. In that respect, I think things are nicely converged in both the shapes and the masses. I think the only way to make significant improvements at this point would be to run additional models targetting low Tmaj/Tmin, but really am not sure it's worth it at this stage.
- I think I'm comfortable calling things at this stage and reporting the results below. This would mean that the new parameters would be:
  - Mbh = 3.0+/-0.3 billion Msun
  - M/L = 2.49 +/- 0.06
  - M15 = 0.99e12 Msun
  - u = 0.998+/- 0.001
  - p = 0.854 (+0.003)(-0.004)
  - q = 0.833 (+0.002)(-0.001)
Now, here are some diagnositcs starting with the 1d panels:

All Models	Best Scales (with Scale Colors)	Best Scales (Colored by Grid)

Here's a comparison of the GPR and dynesty results before and after using the most recent set of three scalings:
- Note that these sets of cornerplots use consistent definitons of everything (e.g., all parameters are straight from what's in TriOS), and these are slightly different to the production level plots below which perform the cosmological corrections we've discussed. These were also made using nIter=8 so the 3 sigma region is perhaps a bit messy, but these are improved in the production level plots which I reran with nIter=8.

And lastly, here's an up-to-date "production plot" for the two K=60 cases above (they seem to do the best at resolving Tmaj/Tmin, though still not the entire 3 sigma region). These both use nIter=8 as well. The only difference between the two panels seems to be a 0.1 degree difference in theta/phi.

Without Newest Scalings	With Newest Scalings