Error in estimation - Log density is -Inf

can anybody help me? I am trying to estimate a model and the following message appears:
“There’s probably a problem with the modified harmonic mean estimator.”
Moreover: Log density is -Inf.

No, this usually means there is a problem with the mode so that the normal approximation to the posterior density does not work. Have a look at the mode_check-plots. If there were stochastic singularity, you would not get that far.

Dear Professor Johannes,

thank you very much for your explanation.
How can I solve this problem?
I already tried several mode_compute options.
Is it possible for a model to run on an older version of matlab and not run on a newer version? Even though it’s the same version of dynare?

Your help is very important for my research.

Thank you very much for your help

  1. As I said, investigate the mode_check plots for problems.
  2. Yes, Matlab versions can make a difference, but that is really rare.

Dear Professor,

thank you again.
I am sending attachment the graphics. Could you help me understand what is going on?mode_check1.fig (37.3 KB) mode_check2.fig (37.5 KB)

You clearly did not find the mode. Given the size of the posterior density, there is most probably something wrong with your observation equation/data treatment

Thank you very much Prof. Pfeifer.
But, I dont know how to interpret the mode_check plots for problems. where can i learn about it?

The line for the mode should be at the peak. That is clearly not the case for most parameters.