# Question about estimation

I have a question about estimation. How is the file with quarterly panel data constructed? I use excel file with five columns for five variables, as well as two columns for country and quarter, but I get error message.

error: findobj: properties and options must be strings

Data.xls (79.5 KB)

What exactly are you trying to do?

I’m trying to do an estimation based on panel data from real world, using a type of Metropolis-Hastings algorythm - Random Walk Metropolis.
Model.mod (2.6 KB)

But how do you want to incorporate the panel structure into DSGE estimation?

I’m not sure. I linearize the non-linear system of equations and then solve the system using Blanchard Kahn method for system reduction. Finally, I want to do an estimation. First, I set priors, second, to estimate model parameters, I use likelihood function “Kalman filter” and finally, to obtain a posterior distribution, I use Markov Chain Monte Carlo simulations (Metropolis-Hastings algorythm).

What I mean is: what makes it a panel? Do you simply want to pool all observations and treat them as one unit?

Oh, I want to treat observations as three units, i.e. countries Slovenia, Austria and Italy.

But how is estimation of a DSGE model with a panel data structure supposed to work?

I don’t have any experience with estimation of DSGE models. I’m trying to replicate the structural parameter stability tests, as conducted by Mr Oke Röhe (a comparison of parameter instability between countries).

If panel data estimation doesn’t work, I could pool observations together I guess, and then do the instability tests for each country separatly?

What exactly is the reference?

I am attaching short version and here is full PhD thesis: https://d-nb.info/1026165547/34 (around page 62/262)

Jerger Roehe 2011.pdf (261.6 KB)

But that paper estimates the model for one country at a time.

Okay, thank you, so how do I construct Excel file for one country-estimation? Using file in attachment, I still get error message:

error: findobj: properties and options must be strings

Data.xls (55.5 KB)
Model.mod (2.5 KB)

What is the full error message? I am getting

Estimation: the ‘estimated_params’ block is mandatory (unless you are running a smoother)

I obtain a table of “Approximated coefficients of correlation” and charts as output, followed by just this line:

`error: findobj: properties and options must be strings`

I’m running Dynare version 4.6.1 on Octave version 5.2.0. I attached print screen.

That looks like an Octave issue. Can you try Dynare 5.4 with Octave 8.1?

I tried, now I obtain yet another error:

error: set: invalid handle (= 1)
error: called from
print at line 492 column 8
dyn_saveas at line 42 column 5
stoch_simul at line 353 column 25
driver at line 375 column 27
dynare at line 278 column 5

Did you close a figure that was popping up?

Yes, otherwise everything freezes…

So that explains the error message. How long did you wait to judge that everything freezes?