Endless estimation

Hi all,
I want to estimate a two-country DSGE model in Dynare but even with mh_replic=10 it takes forever (as in: since yesterday afternoon and still running…)
I use steady_state_model so that can’t be the issue. Do you have other ideas what could be the (or an obvious) issue? Any help is greatly appreciated.

nestedCES.mod (22.1 KB)

dataforestimation.mat (7.4 KB)

The steady state file is missing. Did mode-finding conclude? And which Dynare version are you using?

I use the steady_state_model block in the .mod-file instead of a separate steady-state-file. Is that inefficient?
I’m using Dynare 4.4.3.
What exactly do you mean by ‘did mode-finding conclude’?
The estimation was ‘climbing the hill’ already but it just doesn’t seem to end :frowning:

  1. The ToT_N_steady_state you use is missing.
  2. So your model is still in the process of mode-finding. Your model is big and mode_compute=6 is very slow. That should explain it. You may want to use Dynare 4.5 as the defaults for mode_compute=6 have changed. Is there a reason you cannot use a faster mode-finder?

I use the latest stable version now with mode_compute=1 but get the error message:

Error using chol
Matrix must be positive definite.
Error in posterior_sampler_initialization (line 84)
d = chol(vv);
Error in posterior_sampler (line 59)
[ ix2, ilogpo2, ModelName, MetropolisFolder, fblck, fline, npar, nblck, nruns,
NewFile, MAX_nruns, d, bayestopt_] = …
Error in dynare_estimation_1 (line 443)
Error in dynare_estimation (line 105)
Error in nestedCES (line 795)
Error in dynare (line 223)
evalin(‘base’,fname) ;

The mode check plots also scream trouble -.-
Any further hints what could be the issue here?

ToT_N_steady_state.m (642 Bytes)
C_steady_state.m (351 Bytes)
C_star_steady_state.m (507 Bytes)

  1. Some of your data seems to have a strong seasonal pattern.
  2. Your observation equations are clearly wrong. All inflation rates in the model are mean 0, but in the data they are not. Please take a look at Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models”

Thanks a lot Johannes,
yeah that was a stupid mistake with the observation equations. I removed the seasonalities in the inflation rates and let the estimation run. The key parameters of interest for me are alpha and gamma.
Dynare is not able to run the estimation with mode_compute=1 (hessian is not positive) so I used mode_compute=6 to check whether this one works at least. The procedure found the mode and gave me the attached results (i included the plots in the .zip file). I’m pretty puzzled by how tight the posterior distributions are. What could be an obvious issue here? These results are clearly fishy.
Also: I am not so sure how to treat R_obs in the data. We have the downward trend due to the crisis in 2008. Would I have to remove this trend?
Sorry for any inconveniences!

<a class=“attachment” href="/uploads/default/original/2X/a/a89028ca6583865452358ab9c9estimation plots.zip (115.9 KB)
a998b354edb762.m">C_star_steady_state.m (507 Bytes)
ToT_Nstar_steady_state.m (578 Bytes)
C_steady_state.m (351 Bytes)
nestedCES.mod (22.1 KB)
dataforestimation.mat (7.4 KB)