weird results

[marcio_fr]

Hi,

I'm trying to estimate a DSGE model based on Gali and Monacelli (2005) with bayesian techniques, but I'm getting some weird results. When I give a positive shock in the monetary policy rule, the bayesian impulse response function shows a fall in the interest rate! Other strange result is the kernel in the rhoa parameter (persistence parameter in the technical progress), in the checkplot graphic. I used the script available on Macromodel and made some changes in the taylor rule and to run bayesian estimation (attached with the graphics). Those problems could be due to some misspecification in the model? Or it's a problem in the estimation commands?

Best, Márcio.

[jpfeifer]

Please do not use the Macro model base version for estimation. It does not correctly take parameter dependencies into account. This will screw up estimation. A correct version can be found at https://github.com/JohannesPfeifer/DSGE_mod/tree/master/Gali_Monacelli_2005
Also, there is the issue how you specify your observation equation, see http://www.dynare.org/phpBB3/viewtopic.php?f=1&t=6720

[aaronsun]

Please do not use the Macro model base version for estimation. It does not correctly take parameter dependencies into account. This will screw up estimation. A correct version can be found at https://github.com/JohannesPfeifer/DSGE_mod/tree/master/Gali_Monacelli_2005
Also, there is the issue how you specify your observation equation, see http://www.dynare.org/phpBB3/viewtopic.php?f=1&t=6720

Hi Johannes,

Your mod file for Gali and Monacelli (2005) runs well but model diagnostics show that there is 3 collinear relationships between prices and the exchange rate. Of course, according to the model assumption, they are perfectly correlated. Is the collinearity screwing up estimation? All matrices would be singular in this case.

Thanks,
Aaron

[jpfeifer]

No, this is not a problem. The collinearity comes from the unit roots, which are correctly handled via the diffuse filter.