platon
platon copied to clipboard
Plotting the results
Hi, I've been using PLATON with different sets of free parameters. For some of these runs, despite the posteriors being almost Gaussian, the unbinned best fit solution and confidence bands are way out from the binned best fit. Do you have any idea why?
Hi Eduardo,
Could you send me an example retrieval, so that I can reproduce the problem?
Cheers,
Michael
From: EduardoCristo @.> Sent: Sunday, November 27, 2022 6:29 AM To: ideasrule/platon @.> Cc: Subscribed @.***> Subject: [ideasrule/platon] Plotting the results (Issue #49)
Hi, I've been using PLATON with different sets of free parameters. For some of these runs, despite the posteriors being almost Gaussian, the unbinned best fit solution and confidence bands are way out from the binned best fit. Do you have any idea why?
— Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/ideasrule/platon/issues/49__;!!BpyFHLRN4TMTrA!836OKkcPITv551QeYKjN2S09YpCT_D0F4_YBivAdimacWgsVTr7I7yUyDpWeA0SuzvfLN9r20eGhRlTq68XCS95PL1_i$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AANE5IN4CJKPRID5CQXTVO3WKNH33ANCNFSM6AAAAAASMP3G6I__;!!BpyFHLRN4TMTrA!836OKkcPITv551QeYKjN2S09YpCT_D0F4_YBivAdimacWgsVTr7I7yUyDpWeA0SuzvfLN9r20eGhRlTq68XCS9gUjK8F$. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hi Michael, I can share with you the outputs but not the data I'm fitting since it is protected right now. What I can say is that the model seems to explain reasonably the transmission spectrum as most points are at 1-sigma from it. The fit_info is the following:
#create a FitInfo object and set best guess parameters fit_info = retriever.get_default_fit_info( Rs=0.766* R_sun, Mp=1.138 * M_jup, Rp=R_guess, T=T_guess, logZ=1.08, CO_ratio=0.66, log_cloudtop_P=4, log_scatt_factor=0, scatt_slope=4, error_multiple=1, T_star=4969,T_spot=3719,spot_cov_frac=0.05)
#Add fitting parameters - this specifies which parameters you want to fit fit_info.add_gaussian_fit_param('Rs', 0.013R_sun) fit_info.add_gaussian_fit_param('Mp', 0.025M_jup) fit_info.add_gaussian_fit_param('Rp', 0.00370.766R_sun) fit_info.add_uniform_fit_param('T',800, 2000) fit_info.add_uniform_fit_param("log_scatt_factor", 0, 1) fit_info.add_gaussian_fit_param("logZ", 0.23) fit_info.add_uniform_fit_param("scatt_slope", 2,30) fit_info.add_uniform_fit_param("log_cloudtop_P", -0.99, 5) fit_info.add_gaussian_fit_param("CO_ratio", 0.09) fit_info.add_uniform_fit_param("error_multiple", 0.5, 3) fit_info.add_uniform_fit_param("spot_cov_frac", 0.,0.5) fit_info.add_uniform_fit_param("T_spot",3000,4719)
Hi Eduardo,
The transit depth on that plot is 17%, which is far bigger than for any known exoplanet (except those orbiting white dwarfs and the like). Are you sure that you fed the transit depths to PLATON, and not Rp/Rs?
Cheers,
Michael
From: EduardoCristo @.> Sent: Monday, November 28, 2022 11:06 AM To: ideasrule/platon @.> Cc: ideasrule @.>; Comment @.> Subject: Re: [ideasrule/platon] Plotting the results (Issue #49)
Hi Michael, I can share with you the outputs but not the data I'm fitting since it is protected right now. What I can say is that the model seems to explain reasonably the transmission spectrum as most points are at 1-sigma from it. The fit_info is the following:
#create a FitInfo object and set best guess parameters fit_info = retriever.get_default_fit_info( Rs=0.766* R_sun, Mp=1.138 * M_jup, Rp=R_guess, T=T_guess, logZ=1.08, CO_ratio=0.66, log_cloudtop_P=4, log_scatt_factor=0, scatt_slope=4, error_multiple=1, T_star=4969,T_spot=3719,spot_cov_frac=0.05)
#Add fitting parameters - this specifies which parameters you want to fit fit_info.add_gaussian_fit_param('Rs', 0.013R_sun) fit_info.add_gaussian_fit_param('Mp', 0.025M_jup) fit_info.add_gaussian_fit_param('Rp', 0.00370.766R_sun) fit_info.add_uniform_fit_param('T',800, 2000) fit_info.add_uniform_fit_param("log_scatt_factor", 0, 1) fit_info.add_gaussian_fit_param("logZ", 0.23) fit_info.add_uniform_fit_param("scatt_slope", 2,30) fit_info.add_uniform_fit_param("log_cloudtop_P", -0.99, 5) fit_info.add_gaussian_fit_param("CO_ratio", 0.09) fit_info.add_uniform_fit_param("error_multiple", 0.5, 3) fit_info.add_uniform_fit_param("spot_cov_frac", 0.,0.5) fit_info.add_uniform_fit_param("T_spot",3000,4719)
— Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/ideasrule/platon/issues/49*issuecomment-1329444662__;Iw!!BpyFHLRN4TMTrA!7TUSHw8dtrjs9lEJWST9LJ28pIl8OETIg-Kd_d-GsPC_KD6MYEVe8N3YYRnjcbL2nXAJUEMvGkP7yhvatExyE2nquSxy$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AANE5IMDHV4JTW5YTQ7BC53WKTRANANCNFSM6AAAAAASMP3G6I__;!!BpyFHLRN4TMTrA!7TUSHw8dtrjs9lEJWST9LJ28pIl8OETIg-Kd_d-GsPC_KD6MYEVe8N3YYRnjcbL2nXAJUEMvGkP7yhvatExyE9JE_YCa$. You are receiving this because you commented.Message ID: @.***>
Hi Michael, My original transmission spectrum is measured in Rp/R* and I converted it to transit depths to feed to PLATON. I've converted it back again to build the plot I sent you. The y-axis label has a typo indeed, it should be Rp/R* but there is no mistake in the data input.
Thanks, Eduardo.
Great, thanks! I'm glad the transit depths are not a problem.
It's strange that even the unbinned spectrum is so smooth. There are usually lots of lots of molecular absorption lines all over the place. Do you know what happened to them?
From: EduardoCristo @.> Sent: Monday, November 28, 2022 1:52 PM To: ideasrule/platon @.> Cc: ideasrule @.>; Comment @.> Subject: Re: [ideasrule/platon] Plotting the results (Issue #49)
Hi Michael, My original transmission spectrum is measured in Rp/R* and I converted it to transit depths to feed to PLATON. I've converted it back again to build the plot I sent you. The y-axis label has a typo indeed, it should be Rp/R* but there is no mistake in the data input.
Thanks, Eduardo.
— Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/ideasrule/platon/issues/49*issuecomment-1329672515__;Iw!!BpyFHLRN4TMTrA!4i7ymSOIt-yI1qnFjhJ5SnTmqg1vsRbgiZj7eABRaoOK08u7Wfl_shHJhXhZPyWPc4X3lRu3uc0Zoz1Y8YmHnxKWiJ60$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AANE5IJQ3LKNQRZFU3AVSSTWKUERXANCNFSM6AAAAAASMP3G6I__;!!BpyFHLRN4TMTrA!4i7ymSOIt-yI1qnFjhJ5SnTmqg1vsRbgiZj7eABRaoOK08u7Wfl_shHJhXhZPyWPc4X3lRu3uc0Zoz1Y8YmHn4KhpXhJ$. You are receiving this because you commented.Message ID: @.***>
Hi Michael, the transmission spectrum that I feed into PLATON is broadband and dominated by Rayleigh scattering. As such, it is somewhat expected to be smooth in these wavelength regions. When I fit with Rayleigh scattering alone the thing converges very well. This type of plot happens mostly when I fit for the effect of the unnoculted stellar spots.
Cheers, Eduardo.
This error is due to the stellar correction factors (due to spots) not being applied to the unbinned data and the 1 sigma constraints. A fix to this issue has been proposed and will be merged into the main branch shortly.