meeting 2023 08 15 n57 - JacobPilawa/TriaxSchwarzschild_wiki

Context

This bullet point contains some diagnostic plots trying to reproduce the smoother-looking posteriors from our old thinned test bullet points. I ended up trying a few different things to reproduce those results, including varying the assumed error on the chi2 and the number of iterations used in dynesty.
- I also experimented a bit with the thinning routine, and how many models we are keeping in each confidence level interval.
- Note that these are all compared with the most naive test where we feed in all models with a simple K = 40 cutoff. To do this comparison, the thinned models are all cutoff at a confidence level of 5.04 which correpsonds to K=40 cutoff in 6d.
- I'm having a hard time reproducing exactly the same non-bimodal posteriors that we had in the previous tests.

Naive Case

	niter=1	niter=2
err=0.5	images/230815/n57_K40_nu1.5_nsamples1_err0p5-1.png	images/230815/n57_K40_nu1.5_nsamples1_err0p75-1.png
err=0.75	images/230815/n57_K40_nu1.5_nsamples2_err0p5-1.png	images/230815/n57_K40_nu1.5_nsamples2_err0p75-1.png
err=1.0		images/230815/n57_K40_nu1.5_nsamples2_err1p0-1.png

Thinning and Keeping all within 1 sigma, out to 5 sigma (K=40) (154 in 1sigma)

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/realization_0_nu1.5-1.png	images/230815/realization_1_nu1.5-1.png	images/230815/realization_2_nu1.5-1.png	images/230815/realization_3_nu1.5-1.png	images/230815/realization_4_nu1.5-1.png	images/230815/realization_5_nu1.5-1.png	images/230815/realization_6_nu1.5-1.png	images/230815/realization_7_nu1.5-1.png	images/230815/realization_8_nu1.5-1.png	images/230815/realization_9_nu1.5-1.png

Thinning and keeping all within 1 sigma out to max points and assuming error =0.5

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/full_max_realization_0_nu1.5-1.png	images/230815/full_max_realization_1_nu1.5-1.png	images/230815/full_max_realization_2_nu1.5-1.png	images/230815/full_max_realization_3_nu1.5-1.png	images/230815/full_max_realization_4_nu1.5-1.png	images/230815/full_max_realization_5_nu1.5-1.png	images/230815/full_max_realization_6_nu1.5-1.png	images/230815/full_max_realization_7_nu1.5-1.png	images/230815/full_max_realization_8_nu1.5-1.png	images/230815/full_max_realization_9_nu1.5-1.png

Thinning and Keeping all within 1 sigma, out to 5 sigma (K=40) and assuming error = 1.0)

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/noisy_full_max_realization_0_nu1.5-1.png		images/230815/noisy_full_max_realization_1_nu1.5-1.png		images/230815/noisy_full_max_realization_2_nu1.5-1.png		images/230815/noisy_full_max_realization_3_nu1.5-1.png		images/230815/noisy_full_max_realization_4_nu1.5-1.png

Thinning and keeping 1/2 within 1 sigma out to 5 sigma (K=40) (~79 in 1sigma)

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/thinned_realization_0_nu1.5-1.png	images/230815/thinned_realization_1_nu1.5-1.png	images/230815/thinned_realization_2_nu1.5-1.png	images/230815/thinned_realization_3_nu1.5-1.png	images/230815/thinned_realization_4_nu1.5-1.png	images/230815/thinned_realization_5_nu1.5-1.png	images/230815/thinned_realization_6_nu1.5-1.png	images/230815/thinned_realization_7_nu1.5-1.png	images/230815/thinned_realization_8_nu1.5-1.png	images/230815/thinned_realization_9_nu1.5-1.png

Thinning and keeping 1/2 within 1 sigma out to max points (~79 in 1sigma)

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/full_thinned_realization_0_nu1.5-1.png	images/230815/full_thinned_realization_1_nu1.5-1.png	images/230815/full_thinned_realization_2_nu1.5-1.png	images/230815/full_thinned_realization_3_nu1.5-1.png	images/230815/full_thinned_realization_4_nu1.5-1.png	images/230815/full_thinned_realization_5_nu1.5-1.png	images/230815/full_thinned_realization_6_nu1.5-1.png	images/230815/full_thinned_realization_7_nu1.5-1.png	images/230815/full_thinned_realization_8_nu1.5-1.png	images/230815/full_thinned_realization_9_nu1.5-1.png

Thinning and keeping 1/2 within 1 sigma out to max points (~79 in 1sigma) and assuming error = 0.75

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/noisy_full_thinned_realization_0_nu1.5-1.png	images/230815/noisy_full_thinned_realization_1_nu1.5-1.png	images/230815/noisy_full_thinned_realization_2_nu1.5-1.png	images/230815/noisy_full_thinned_realization_3_nu1.5-1.png	images/230815/noisy_full_thinned_realization_4_nu1.5-1.png	images/230815/noisy_full_thinned_realization_5_nu1.5-1.png	images/230815/noisy_full_thinned_realization_6_nu1.5-1.png	images/230815/noisy_full_thinned_realization_7_nu1.5-1.png	images/230815/noisy_full_thinned_realization_8_nu1.5-1.png	images/230815/noisy_full_thinned_realization_9_nu1.5-1.png

Thinning and keeping 1/2 within 1 sigma out to max points (~79 in 1sigma) and assuming error = 1.0

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/noisier_full_thinned_realization_0_nu1.5-1.png	images/230815/noisier_full_thinned_realization_1_nu1.5-1.png	images/230815/noisier_full_thinned_realization_2_nu1.5-1.png	images/230815/noisier_full_thinned_realization_3_nu1.5-1.png	images/230815/noisier_full_thinned_realization_4_nu1.5-1.png	images/230815/noisier_full_thinned_realization_5_nu1.5-1.png	images/230815/noisier_full_thinned_realization_6_nu1.5-1.png	images/230815/noisier_full_thinned_realization_7_nu1.5-1.png	images/230815/noisier_full_thinned_realization_8_nu1.5-1.png	images/230815/noisier_full_thinned_realization_9_nu1.5-1.png

Thinning and keeping only ~50 models in each confidence region

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/very_thin_max_realization_0_nu1.5-1.png	images/230815/very_thin_max_realization_1_nu1.5-1.png	images/230815/very_thin_max_realization_2_nu1.5-1.png	images/230815/very_thin_max_realization_3_nu1.5-1.png	images/230815/very_thin_max_realization_4_nu1.5-1.png	images/230815/very_thin_max_realization_5_nu1.5-1.png	images/230815/very_thin_max_realization_6_nu1.5-1.png	images/230815/very_thin_max_realization_7_nu1.5-1.png	images/230815/very_thin_max_realization_8_nu1.5-1.png	images/230815/very_thin_max_realization_9_nu1.5-1.png

Thinning and keeping only ~50 models in each confidence region with an assumed error of 1.0

r1	r2	r3	r4	r5	r6	r7	r8	r9	r10
images/230815/noisy_very_thin_max_realization_0_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_1_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_2_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_3_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_4_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_5_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_6_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_7_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_8_nu1.5-1.png	images/230815/noisy_very_thin_max_realization_9_nu1.5-1.png