meeting 2025 03 21 n57 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

Context

  • The bullet from yesterday was getting quite messy, so I've moved relevant plots to this page with a clearer explanation of what each case is showing.
    • Note that in all the plots below, I've subtrated the median V from all data for consistency. This had been slightly inconsistent in the previous page with the top, first column set of plots not having the median subtracted, but the rest of the set of plots being median subtracted.

MC Fits

  • The left column here is comparing the kinematics from the old code vs new code for the same set of settings. The settings used are:

    • adeg = 0
    • bias = 0.2
    • correct Barth stars
    • nMc = 100

  • The right column is showing the new code kinematics vs. the kinematics currently used in our minimizations. The kinematics currently used in our minimizations were run with very different settings, using different values of:

    • bias = 0.0
    • off-by-one Barth stars
    • these still should have been using nMc = 100 and adeg=0, however.
    • Despite these different settings, there is still quite good agreement between the datasets.

  • The second row of plots here show the same data but without errorbars for clarity.

  • Summary/Takeaways:

    • The old and new code for the same settings are mostly in agreement (left column of plots), though the high sigma points are a bit elevated using the new code, and the even moments appear to be ever so slightly higher in the new code, too. The vast majroity of the data fall within 1 sigma of the 1-to-1 line, however.
    • The new code kinematics vs. current code kinematics agree with one another exceptionally well I think, given that they use totally distinct stellar templates and different bias values. I think that the level of agreement here is basically on par with what we saw for N315, with slight tilts in the one-to-one plots but nothing totally off base.
Old pPXF Code vs. New pPXF Code New Code Kinematics vs. current kinematics used in the minimization
with errors [images/250321/caseI_vs_caseZ.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250321/caseZ_vs_old_old_old_kinematics.png)
without errors [images/250321/caseI_vs_caseZ_no_errors.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250321/caseZ_vs_old_old_old_kinematics_no_errors.png)

Preliminary Fits

  • We then wanted to look at the preliminary fits rather than the fits after MC realizations. These plots are below:

  • The left column here is again showing the old code vs. new code for the same input settings, but this time I've plotted the preliminary fits alone. Essentially the preliminary version of the left column from above.

  • The right column is showing the newest pPXF fitting code preliminary fits vs. the current kinematics plotted without their errorbars since I don't have the preliminary fits for this anymore. I'll run this case and hopefully have that ready for the meeting.

  • Results/Takeaways:

    • I don't think either set of plots here are inconsistent with the conclusions from above. It seems like there is a decent level of agreement (thought slight offsets) with the new vs. old code, with the new code being slightly inflated in the even moments.
    • The old kinematics vs. new kinematics are still in decent agreement, again surprising given the different settings used.
Old pPXF Code vs. New pPXF Code New Code Kinematics vs. current kinematics used in the minimization
[images/250321/caseI_vs_caseZ_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250321/caseZ_prelim_vs_old_old_old_kinematics.png)

Reproducing Kinematics Currently Used in Minimization with the new code

  • One thing we wanted to check was that we can broadly reproduce the original kinematics currently used in the minimization with the new code, and I think I was able to do so at a comparable level to N315. On this page, there's a plot labelled 'Reproducing My Original Kinematics' which should correspond to the test I have below. This is basically taking the new code and using (incorrect Barth stars + bias = 0.0 + adeg = 0 + nMc = 100 + the same starting guess, etc)
  • Results:
    • It seems like we can basically reproduce what we have for the current kinematics, with the caveat that there are a few bins which are certainly outliers. I can look at these more carefully but I was happy to see that the vast majroity of our data scatter around the 1-to-1 line here after so many changes to the pPXF fitting routine:
    • Note that I don't have the preliminary fits for the original kinematics used in the minimization, so the best I can do is plot the comparison of the MC moments, but I do have a version showing the preliminary fits from the new code vs. the median over the 100 MC realizations without error bars for clarity:
Reproducing the original kinematics used in the minimization Prelim reproduction vs. Kinematics currently used in the minimization
[images/250321/caseW_vs_old_old_old_kinematics.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250321/caseW_prelim_vs_old_old_old_kinematics.png)

Quick Mid Meeting Update

  • x-axis here should be me attempting to reproduce my original kinematics with the new code
  • y-axis here should be the newest set of kinematics with the updated code + correct set of parameters
Plot
images/250321/test3.png