Changes in mode_compute=4 between Dynare 4.4.3 and 4.5.3?

Estimation ML/Bayesian estimation
1

Hi guys,

I want to re-estimate a large-scale DSGE model from a colleague at another institution. On his computer with MATLAB 2018b/Dynare 4.4.3, the following estimation command works without warnings or error messages:

estimation(datafile=FILENAME, mode_file=FILENAME, order = 1, mode_compute=4, mh_replic=0, plot_priors = 0, mode_check) ger_gdp eur_gdp;

If I run the same code with the same data file and the same mode file on my computer with MATLAB 2019a/Dynare 4.5.3, I find exactly the same modes, but the posterior variances are NaN and I get the message:
“POSTERIOR KERNEL OPTIMIZATION PROBLEM!
(minus) the hessian matrix at the “mode” is not positive definite!
=> posterior variance of the estimated parameters are not positive.
You should try to change the initial values of the parameters using
the estimated_params_init block, or use another optimization routine.”

Have there been any changes to how mode_compute=4 works between Dynare 4.4.3 and 4.5.3? I am wondering whether Dynare 4.4.3 erroneously did not point to any problems or whether there is a bug with mode_compute=4 in Dynare 4.5.3.

Thanks for any hints on this issue!

2

Could you please provide me with the files to investigate the issue.

3

I am sorry but the authors whose original code I am using would rather not like to see their code posted here.

I have circumvented the problem by using mode_compute=5 together with optim=(‘Hessian’, 2). This gives me modes and variances that are very close to the authors’ results.

Thanks a lot anyway.

4

@dgroll You don’t need to post the code here. You could send me a PM or to my email address.