Page 1 of 1

check_plots

PostPosted: Wed Aug 28, 2013 1:21 pm
by ger163609
Hi!

I have a problem with the check plots from the check_mode command: The maximum the optimizer finds is not the same as indicated by the posterior (blue line). How can I fix this problem?
By tighter priors? A different optimizer?

Thanks in advance for your help!

Re: check_plots

PostPosted: Mon Sep 02, 2013 3:17 pm
by jpfeifer
Tighter priors will most probably not help as they do not give you more curvature. A different mode-finder should be more successful. Try 8 or 9.

Re: check_plots

PostPosted: Tue Sep 03, 2013 4:01 pm
by ger163609
Thanks jpfeifer!

I tried mode_compute=9 and now the MCMC statistics are fine.

Anyway, do you know whether the MCMC draws and the posterior distributions are stored somewhere? I cannot find them in the workspace. I need to plot the posteriors and the Brooks/Gelman statistics for each parameter seperately, i.e. not jointly in this 3x3 subplot array.

Thanks.

Re: check_plots

PostPosted: Tue Sep 03, 2013 5:01 pm
by jpfeifer
In this case it is probably easiest to hack the McMCDiagnostics_core.m file.

Re: check_plots

PostPosted: Thu Sep 05, 2013 2:20 pm
by ger163609
thanks!