error: MJDGGES returns the following error code: 1176
error: called from:
error: c:\dynare\4.2.2\matlab\print_info.m at line 36, column
error: c:\dynare\4.2.2\matlab\check.m at line 51, column 5
error: c:\Daniel_QPM\gpm_euperiph_apr2012\soe1bayes_irfs.m at
error: c:\dynare\4.2.2\matlab\dynare.m at line 120, column 1
This is trying to run stoch_simul after adding some more stationnary shock processes to a file which before was running without any problems. As far as I can tell this has nothing to do with existence/indeterminacy of solution issues or steady state solution issues, since I didn’t get any messages related to this. Any help would be most welcome.
Also, is there any documentation of the possible MJDGGES error codes and their meaning.
I couldn’t find anything in the manual, I can’t even see the contents of MJDGGES and it’s very frustrating to try to debug an error like this which doesn’t give any clear hint on what can be wrong.
I’ve attached the mod file, ss file and extra data file you need to try to run this.
data3ctry.m (217 KB)
soe1bayes_irfs_steadystate.m (35.7 KB)
soe1bayes_irfs.mod (275 KB)
If you run
Dynare says that there are 68 collinear equations. Hence, try to fix the model.
I’ll try to have a more detailed look into possible bugs when I have time to get back to it. This model setup worked without any problems for other multi country models, so hopefully it’s a really silly error.
But in the meantime, what do you mean by collinear equations? Equations that are linearly dependent so the system can be reduced to one with fewer equations than the number of variables? If so, shouldn’t this have violated the rank or the basic determinant condition for solvability (det(Az-B)~=0 for some z to use Klein/King and Watson notation), and isn’t there a more explicit warning to that effect the same way the check command warns you if there are too many stable/unstable eigenvalues or if there 0/0 or unit roots eigenvalues?
Yes, I mean that your model has not full rank. In theory, you are correct that this should be easily detectable. But remember, we are dealing with finite precision computing for infinite precision real numbers. Hence, theoretically, an eigenvalue should be 0 as you said. But testing for it is not easy as we are dealing with huge matrices in your case and there is always rounding involved. Your model may be rank-deficient. But it could also be that due to the size of your model, the collinearity check is not precise enough and gives a false warning.
Ok, thanks a lot for the clarification.
Just got back to the file and everything’s now solved.
For future reference for stuff to check for if anyone gets this sort of message:
Turns out I accidentaly typed in the same equation twice (a risk you take when using copy/paste to produce multi country models).