Dynare forums

Dear all,

I am estimating the standard deviations of two structural shocks and their correlation in a simple RBC model.
I am running identification analysis tests via the command

Code: Select all

identification;

and am getting two error messages.

1) When I run the .mod file and apply the identification test on the standard deviations only or on all paramaters
here is the error message that I get

Code: Select all

Error using  ./
Matrix dimensions must agree.

Error in identification_analysis (line 222)
        ide_strength_J_prior(indok) = (1./(normaliz(indok)'./normaliz1(indok)'));

Error in dynare_identification (line 280)
    [idehess_point, idemoments_point, idemodel_point, idelre_point, derivatives_info_point, info] = ...

Error in rbc_estim (line 219)
dynare_identification(options_ident);

Error in dynare (line 180)
evalin('base',fname) ;

2) If I run the identification test on the correlation between structural shocks only the reported error
message is

Code: Select all

Error using  .*
Matrix dimensions must agree.

Error in identification_analysis (line 224)
        deltaM_prior = deltaM.*abs(normaliz1');

Error in dynare_identification (line 280)
    [idehess_point, idemoments_point, idemodel_point, idelre_point, derivatives_info_point, info] = ...

Error in rbc_estim (line 218)
dynare_identification(options_ident);

Error in dynare (line 180)
evalin('base',fname) ;

The identification test works fine if it is applied to parameters not estimated via one of Dynares built-in
functions (as, in this case, stderr or corr). Help on the issue is greatly appreciated. I attach my .mod file.

Many thanks in advance,
Francis

Please see http://www.dynare.org/phpBB3/viewtopic.php?f=1&t=5135

Thank you Mr. Pfeifer.

Could not find any forum threads concerning the first error which appears when including standard deviations. As far as I know the identification command should work for those.

Many thanks again!

It does not work for measurement errors. Standard errors should work. Please provide the file to replicate the issue.

It is the file I attached above with the correlation commented out from the
estimation parameters. I attach the orginal file again with this modification.

Thanks!

This is a bug. Thanks for pointing it out. It comes from only testing the identification of exogenous variables' standard deviations. If you add

Code: Select all

sigma , 1,0, 5;


to the estimated_params-block, it works.

Good to know! Many thanks for your kind feedback.

Best,
Francis