Page 1 of 1

Schur decomposition failed

PostPosted: Thu Mar 27, 2014 5:36 pm
by economist
Hello,
Im having trouble with the Schur decomposition failing for my code. The steady state is correctly calculated and the residuals of all equations are exactly zero. I can't seem to find the problem. I'm posting the .mod file and the steady state auxiliary file. Any help will be much appreciated.
Thank you!

Re: Schur decomposition failed

PostPosted: Thu Mar 27, 2014 6:07 pm
by kyri82
If I read your results correctly you have negative values at the steady-state. Are you sure this is correct?

K

Re: Schur decomposition failed

PostPosted: Thu Mar 27, 2014 7:15 pm
by economist
Thank you, K.

I don't think that is the issue. Since I'm log linearizing, those steady state values are logs of the variables. I get the same erro message when I linearize and all steady-state values are positive.

Re: Schur decomposition failed

PostPosted: Thu Mar 27, 2014 7:58 pm
by kyri82
You sure about your parametrisation? Some parametre values sometimes give unit roots. Do you solve the steady-state by hand?

Re: Schur decomposition failed

PostPosted: Thu Mar 27, 2014 8:52 pm
by economist
Yes, I believe so. I solve the steady state in closed-form for all variables. Thank you for looking into this.

Re: Schur decomposition failed

PostPosted: Thu Mar 27, 2014 8:59 pm
by kyri82
You know if you do that you can introduce the log-linearised model directly in dynare. Then you express each variable as a log-deviation from the ss and you do not need to worry about the ss anymore.

Re: Schur decomposition failed

PostPosted: Fri Mar 28, 2014 10:03 am
by jpfeifer
From an experimental version of model_diagnostics:
MODEL_DIAGNOSTICS: The Jacobian of the static model contains Inf or NaN. The problem arises from:

MODEL_DIAGNOSTICS: Derivative of Equation 18 with respect to Variable BD (initial value of BD: 0)
MODEL_DIAGNOSTICS: Derivative of Equation 19 with respect to Variable BF (initial value of BF: 0)

MODEL_DIAGNOSTICS: The problem most often occurs, because a variable with
MODEL_DIAGNOSTICS: exponent smaller than 1 has been initialized to 0. Taking the derivative
MODEL_DIAGNOSTICS: and evaluating it at the steady state then results in a division by 0.