Closed ahallcosmo closed 4 years ago
I could do something like results['samples'] = np.exp(results['samples'])
Yes, that should work with no problems (or at least that's what I do!). No other (meta)data should need to be modified as far as I'm aware, but if you spot weirdness please let me know.
Thanks!
Hi,
I have the results of a Nested Sampling run with a particular set of parameters. If I want to post-process the results to a different parameterisation (e.g., to plot the posterior of exp(x) rather than x), is it just a case of applying the appropriate transform to the samples? So for a 1D problem, with samples from the posterior of x, I could do something like results['samples'] = np.exp(results['samples']) to get a results class entirely equivalent to what I would have got by using exp(x) as a parameter in the sampling rather than x? Or are there any other metadata in the results class that would need to be modified too?
Best wishes, Alex