meeting 2025 08 25 n410 - JacobPilawa/TriaxSchwarzschild_wiki

Context

This page is still a bit of a work-in-progress, but wanted to get notes up this afternoon (August 25).
Following up on last Thursday's brief discussion, in which we decided to move forward with the no mask, sigma clip = 3.0 case for the kinematics for N410.
After making this decision, I wanted to catch our usual diagnostics up to speed (regarding the templates/bias/etc) to verify that we have stable results moving forward.
Takeaways:
- I continue to see very nice agreement between the h6 vs. h8 vs. h10 vs. h12 results now with this new mask + sigma clipping. There doesn't appear to be any systematic trends in our latest set of diagnostics.
- The trend of varying the additive polynomial seems to hold both with no mask + sigma clipping, agains suggesting that we're good with using adeg=0 and mdeg=3 (though I probably should run the specific mdeg test).
- The superspectrum testing is a bit interesting -- it seems like the superspectra is far less sensitive to adeg=0 vs adeg=-1 (at least in terms of sigma), with the results agreeing within a few km/s.
  - N410 still prefers some of the very strange spectra (from the 384 + 3 case); also in some of the preliminary testing, it was a bit hard to get reasonable superspectra fits with the Barth stars alone (need to upload figures for this).
Moving Forward:
- Want to process the subset of superspectra results as we have done in the past; there's a chance that the "subset" results are picking different tempaltes (when we are don't have the original mask).
- Also want to check out the actual differnece between fit and data (to see if there are systematic trends near the center of the galaxy, for example).
- Need to move onto MGE fitting/point symmetrizations.

Diagnostics

I first started by fitting the spectra with our "fiducial" settings, changing only the mask to be blank at first, and adding in the 3.0 sigma clipped pixels as part of the mask afterward. Here's the comparison of changing the number of GH moments:

	h8 vs. h6	h8 vs. h10	h8 vs. h12
one-to-one
diffs

Here's the RMS vs. spectra number for these fits:

h6	h8	h10	h12

And lastly, the fits to the spectra for these cases

h6	h8	h10	h12

testing additive polynomial without a mask

It's worth checking to see if our additive polynomial was "trading off" with the mask in anyway, so I re-ran the test in which I varied the additive polynomial as we fit the spectra. I'm comparing two cases below:
- One case varies the adeg while not using a mask, but allowing the sigma clipping (=3.0) find which pixels are to be removed
- The other case is our old "fiducial" case, in which I use the mask on BOX along with the sigma clipping routine.
In many ways, because the sigma clipping = 3.0 hardly changes which pixels are masked (since the mask from BOX removes a great deal of the already "poorly fit" points), this test is really comparing the effect of the additive polynomial with sigma clipping alone vs. our fidcuial mask (essentially alone).
Note in these figures, the blue points are the BOX mask, and the orange points

Vary adeg (no zoom)

updated superspetrum testing

We had a few other open questions related to the templates now that we've switched away from the BOX mask and toward the 3 sigma clipped mask.
Specifically, we wanted to see if the superspectrum results appreciably changed with the new mask, and also whether or not there is a dependence on the additive polynomial for the superspectrum.

	adeg=0	adeg=-1
No Mask (sigma clip = 3.0)
Fiducial Mask (sigma clip = 3.0)

And here's the top 15 templates selected in each case (note that I'm using the set of 384+3 that we've been playing around with for the last few tests):