Log data density -inf (harmonic mean estimator)

Estimation ML/Bayesian estimation
1

Hello Professor,
I am working on estimating NK model with 16 secotrs.

I am trying to estimate the Calvo parameters with other parameters.
(previously, I just assigned values from the literautre for the Calvo parameters)
With mode_compute=5, I think I got decent mode, which you can check from attached file.

But the problem is the log data density if -inf (from Harmonic mean estimator), and I got following messages:

Estimation::marginal density: There’s probably a problem with the modified harmonic mean estimator.

Besides, the log density from the Laplace approximation is fine as:
Log data density [Laplace approximation] is 9397.546339.

So my question is whether the -inf log data density of Harmonic mean estimator is problem and how can I fix it.
Is it a problem of low draw? I only tried the draw of 30,000 so I’m trying more, but not sure more draws will fix this problem.

I attached mode check and prior and posterior plots, and also my mod file.
Thank you for your help.

Plots.pdf (438.2 KB)

2

The Inf Harmonic mean estimator is due to a non-robust way of computing the log-determinant in Dynare 5.x. It is more robust in the current unstable version, i.e. Dynare 6.x.

3

I really appreciate your help professor.
I will try the version 6.x.

Thank you!