meeting 2025 03 07 n315 pt2 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki
Context
- First, here are the raw cornerplots from our dynesty and GPR routine for what's currently in the paper vs. this newest Grids D + E (best scalings). I am using nu=1.5 and K=100 for both of these cornerplots for consistency, and the parameters quoted here have the same definitions as used in the TriOS code (e.g., cosmological corrections/distance corrections still need to be made), but the two cornerplots are using consistent definitions with one another here.
- Notably the current draft has Tmin and Tmaj 3 sigma contours which close compared to our new cornerplots which only resolve the 2sigma region.
| Current Draft | Updated Cornerplot |
|---|---|
| [images/250307/240110_inner_5_removed_grid_alpha_K100_nu1.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250307/grids_D_and_E_BEST_SCALES_nIter8_grid_alpha_K100_nu1.5-1.png) |
- However since we only ever quote the T's at 1sigma and never show their posteriors, I'm not sure all hope is lost with respect to the shape constraint. I've created an equivalent "production" plot for the newest set of posteriors, and frankly in this way the results are looking very nice with the 3sigma contours in angles and u/p/q being fully closed. Note that these posteriors have the cosmological corrections applied in both cases/have consistent definitions of everything, so these can be compared one-to-one and will be slightly different than what is above due to these corrections.
| Current Draft Cornerplot | Attempted "New" Cornerplot | |
|---|---|---|
| K=100 | [images/250307/240110_inner_5_removed_no_prior_grid_alpha_K100_nu1.5_testing_cutting_weights_cosmo_corrections_M15_change-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250307/grids_D_and_E_BEST_SCALES_nIter8_grid_alpha_K100_nu1.5_testing_cutting_weights_cosmo_corrections_M15_change-1.png) | |
| K=60 | images/250307/grids_D_and_E_BEST_SCALES_nIter8_grid_alpha_K60_nu1.5_testing_cutting_weights_cosmo_corrections_M15_change-1.png |
Testing nu = 0.5
- Focusing on the 3sigma contour issue, I can very slightly improve how things look by switching to nu=0.5, though this is different from what we're currently doing in the draft and I don't think significant enough of a difference for us to bother changing our approach at this point. Here's a quick comparison of nu=0.5 and nu=1.5 for the K=60 and K=100 cases:
| K=60 | K=100 | |
|---|---|---|
| nu=0.5 | [images/250307/grids_D_and_E_BEST_SCALES_quick_test_grid_alpha_K60_nu0.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250307/grids_D_and_E_BEST_SCALES_quick_test_grid_alpha_K100_nu0.5-1.png) | |
| nu=1.5 | [images/250307/grids_D_and_E_BEST_SCALES_quick_test_grid_alpha_K60_nu1.5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250307/grids_D_and_E_BEST_SCALES_quick_test_grid_alpha_K100_nu1.5-1.png) |