Plot the likelihood instead of posterior distribution

Questions
3

Sorry. Maybe I didn’t describe clearly in my original post.

My goal is to find the highly likely values for the parameters \theta of the model.

I think that means I want:

  • Plot of the log-likelihood (logp) as a function of the parameters \theta. The plot will not display anything for the values that the sampler hasn’t visited.
  • Not a histogram of log-likelihood values.
  • Not a histogram of the values of the parameters. That is the posterior distribution.

The traceplot is frequency vs value. I guess that describes the posterior distribution instead of likelihood of the parameters.

I don’t know if I should believe my prior is “weakly-informative”.
I am using Dirichlet instead of uniform prior. I saw that Dirichlet with \alpha = \vec{1} look like a triangle with uniform density. Maybe in that case, the prior is indeed weakly-informative?