0

I have a an MCMC sample file containing a list of points in parameter space. I have the value of the parameters in my model at each point, and the likelihood at each point.

Of course I also have the prior ranges for each parameter.

How can I calculate the model evidence? I mean it's just an integral of the likelihoods over all the parameters with. Can PYMC do it?

Itinerant
  • 31
  • 2
  • 1
    See @Xi'an's excellent answer to my question [here](https://stats.stackexchange.com/questions/209810/computation-of-the-marginal-likelihood-from-mcmc-samples). – lacerbi Feb 15 '17 at 13:05
  • Thanks. Nested sampling seems to be the quickest and easiest method. If I order my points according to likelihood, I obtain more or less a nested sampling output. Or do I? – Itinerant Feb 16 '17 at 00:30

0 Answers0