Can convert the results of mcmc?

[d11122]

Hi, I have a question about mcmc samples.
For example, I performed a mcmc, I performed parameters with 'incl', 'sma','r1','t2' and so on。
now I can get the value and sigmas of M1 (see figure1)

but my question is how to draw the corner figure with M1.
I can't find samples of M1 to plot corner figure.
Does it have to be involved in the mcmc calculation at the beginning?
thanks you!

[kecnry]

You should be able to create a corner plot with b.plot(style='corner') (you will also need to pass the solution tag if you have more than one). If you want to plot a subset of the sampled parameters, you can either set the adopt_parameters parameter in the solution, or pass it as a keyword argument to plot. If you want to plot parameters that were not sampled but are affected by existing constraints, you can do so as well by passing parameters as a list.

If you want to access the arrays and plot manually, see the parameter in the solution tagged with qualifier='sampled' or access them with phoebe.helpers.process_mcmc_chains_from_solution.

You might also find these tutorials/docs helpful:

• emcee tutorial (scroll down to plotting at the bottom)
• distribution propagation
• plot API docs

Hope this helps! Let us know if you run into any problems or questions.

[d11122]

Hi kecnry, thank your response, I can plot the corner figure with not sampled parameters by using phoebe distribution propagation.
For example The figure below shows ,the r2/r1.

But I still can't find where the sample data for r2/r1 is saved? If I want to manually draw parameters that are not sampled but are affected by existing constraints .
I hope you will answer again.
thanks!

[kecnry]

yes, you can pass those same arguments to b.sample_distribution_collection and I think that should give you what you want (note that you'll likely want to pass sample_size and that depending on the cost of the constraint and the passed sample_size, this can get somewhat expensive).

[d11122]

Hi, kecnry, Thanks a lot, I have achieved the result I wanted.