Page 1 of 1

sensitivity analysis error

PostPosted: Wed Nov 20, 2013 3:11 pm
by marco_dp
Hi I'm trying to run a GSA on a large scale DSGE model
the command I used is:

Code: Select all
dynare_sensitivity;


I got this problem:

Code: Select all
All parameter values in the specified ranges are not acceptable!
 
Sensitivity computations stopped: no parameter set provided a unique solution


Moreover, it is repeated the message "SOLVE: maxit has been reached" for 2048 times (equal to the size of MC sample)

can someone help me?
thanks!

Re: sensitivity analysis error

PostPosted: Wed Nov 20, 2013 3:44 pm
by jpfeifer
Looks like Dynare is unable to find the steady state for the parameter draws. Provide better initial values are an analytical steady state.

Re: sensitivity analysis error

PostPosted: Thu Nov 21, 2013 8:51 am
by marco_dp
Ok thanks,

I got a problem also with the identification procedure

Code: Select all
Identification stopped:
The model did not solve for any of 50 attempts of random samples from the prior


the model is a large scale DSGE model (similar to QUEST III) where I do not have the theorical steady-state (in closed form) but only the numerical steady-state (found on the basis of the parameter calibration) both for endogenous and exogenous variables (declared as initval)

Re: sensitivity analysis error

PostPosted: Thu Nov 21, 2013 4:15 pm
by jpfeifer
Seems like the same problem. It looks as if your parameter draw moves away from the calibrated version of the model, the SS is not found anymore. You might need to use a steady state file to narrow the steady state finding down, see http://www.dynare.org/phpBB3/viewtopic.php?f=1&t=5144 or try different steady state finders inside of estimation.

Re: sensitivity analysis error

PostPosted: Thu Nov 28, 2013 3:00 pm
by marco_dp
ok,thanks for your reply

Re: sensitivity analysis error

PostPosted: Tue Dec 10, 2013 9:29 am
by marco_dp
I built an external file for steady-state computing, but when i run identification procedure I got this message

Code: Select all
==== Identification analysis ====
 
Testing prior mean
 
-----------
Parameter error:
The model does not solve for prior_mean with error code info = 19
 
info==19 %! The steadystate routine thrown an exception (inconsistent deep parameters).
-----------
 
Try sampling up to 50 parameter sets from the prior.
 
-----------
Identification stopped:
The model did not solve for any of 50 attempts of random samples from the prior
-----------


How to solve it?

Thanks for the attention

Re: sensitivity analysis error

PostPosted: Tue Dec 10, 2013 10:00 am
by jpfeifer
Please post or send the files.