Mode_check interpretation

Dear all,

I have some question about mode_check graph:

  1. If blue-line and the green-line are overlapped, does it mean that the likelihood domanites the prior?

  2. On the contrary Different curves does mean that the weight attached to a prior is more important that the likelihood (i.e. parameter not-well estimated)?

  3. In the case of a flat curve (as in the case of the parameter rho_r in the attached .pdf file) how do I interpretate this result?

Finally, if I perform a trace_plot, where are store the numerical results?

Thanks for the help!!!
check.pdf (150 KB)

  1. Yes (or that your prior is totally uninformative)
  2. More important would be the wrong word. It shows how the prior updates your likelihood, yielding the posterior. It is not about being well-estimated, you just might not have enough data. For the identification dimension of “well-estimated”, the curvature is important.
  3. For rho_r it is not flat. Look at the scaling. Try zooming into the figure or changing the interval for the plotting (see the manual)

dear jpfeifer, thanks, as usual, for your advice and helpful comments. Thus, based on your own experience, are these paramters ‘well-estimated’?

moreover can you tell me where can I find the store of numerical value of trace_plot?

thanks again!!!

Except for rho, where I cannot judget it, at least the mode finding seems to have worked. The estimates of the posterior distribution cannot be judged from these graphs

1 Like