meeting 2024 12 05 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

Context

  • I've reprocessed the MGE A, B, and C set of 11 models with the corrected triaxnnls binaries so that now the first bin is properly accounted for.

    • Following the changes, it's not obvious to me that the central bin issue was contributing appreciably to the difference between MGEs. There's still a decent amount of scatter comparing MGE A vs. MGE B vs. MGE C (in that they don't perfectly agree 1-to-1).
    • Addtionally, I'm still finding that removing the dummy moments seems to drive part of the difference, as does moving from the kinem chi2 to the NNLS chi2. I don't think that this means we should revert to kinem chi2, but this needs to be investigated a bit further.

  • First, here's the standard 1-to-1 comparing the chi2 values between the MGEs for NNLS/kinem and with/without the dummy moments.

Quantity MGE A vs. B MGE A vs. C
NNLS (No Dummies)
NNLS (with Dummies)
Kinem (No Dummies)
Kinem (with Dummies)

Heatmaps

  • And here are the heatmaps showing the difference in chi2 per bin/per moment for three MGEs. It seems like h6 is driving much of the difference between the models, though I'm looking into this a bit more currently.
Click here to expand heatmaps -- page can be a bit laggy otherwise!
BH Mass/1e9 Msun 2.963 3.2 3.237 2.996 2.529 2.951 3.043 3.026 2.084 2.728 3.091
MGE A - MGE B
MGE A - MGE B

Additional Diagnostics

  • The heatmaps frankly are a bit hard to make sense of other than for broad generalizations of which moments are different between the two models. I've got two sets of plots below that are perhaps a bit more illuminating, in which I'm plotting the model predictions for each moment between two of the MGEs. In one set of plots I've got the standard 1-to-1 plots, and the second set of plots shows histograms of the differences between moments.
BH Mass/1e9 Msun 2.963 3.2 3.237 2.996 2.529 2.951 3.043 3.026 2.084 2.728 3.091
A vs. B
A vs. C
B vs. C
  • And the histograms of differences of model predcitions for h1 through h12. Note that I've got vertical lines at 0, -0.002 and +0.002.
BH Mass/1e9 Msun 2.963 3.2 3.237 2.996 2.529 2.951 3.043 3.026 2.084 2.728 3.091
A vs. B
A vs. C
B vs. C

Comparing Radial Profiles for the Different MGEs

  • I'm trying to figure out why the chi2 might be different between the three MGEs, and instead of looking at the actual values of the moments compared to one another, I thought it might be worthwhile to look at the unfolded radial profiles (in both kinem and NNLS space) to see if there are any systematic reasons why things are different.

  • Below, I'm plotting the radial profiles for the 11 models we've been discussing, and have the moment predictions for each MGE in a different color. MGE A is red, MGE B is blue, and MGE C is green. I've plotted both of the radial profiles in both kinem space and NNLS space (bascically just h1/h2 vs. V/sigma):

    • There's definitely some weird behavior in h2 for the central region, though this also is the case for sigma where the inner ~4 bins are definitely lying above our best fit model predctions. This has been the case virtually our entire time while modeling NGC 315, though I was surprised to see how different things look in h2 space compared to sigma space.

  • Note that the parameters are in the title of each plot. As a quick heads up "dmp1" is rho0 in our models (in code units since I'm pulling directly from the model outputs). I've also rounded the values in the title so there's a very minor difference between the title values and the column names in the actual table.

Type 2.963 3.2 3.237 2.996 2.529 2.951 3.043 3.026 2.084 2.728 3.091
NNLS
Kinem
⚠️ **GitHub.com Fallback** ⚠️