Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
Best
Michel
Michel Juillard michel.juillard@mjui.fr writes:
Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
I would rather drop the linear option from both stoch_simul and model, since I still don't understand its purpose.
1. I agree with Sébastien to drop the option. From my experience it creates more confusion than it's worth. The confusion in this specific case might arise from the manual saying that one should rather put "linear" in model(linear). This suggests that it is indeed equivalent to stoch_simul(linear). However, the manual says that model(linear) sets order=1.
2. By the way: the manual states that using model(linear) spares one from specifying initval. This is not really true as linearity per se implies that there can be constant terms that need to be taken care of. In particular, I have seen a few mod-files in the forum that use a Taylor rule with the target interest rate as a constant. Hence, although the model is linear, the steady state of the interest rate is not 0 and Dynare crashed.
-- Johannes Pfeifer Haußerstr. 29 72076 Tübingen Tel.: +49-(0)7071-6396184 Mobil.: +49-(0)170-6936820 Germany
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Sébastien Villemot Gesendet: Donnerstag, 5. Juli 2012 18:15 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Michel Juillard michel.juillard@mjui.fr writes:
Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
I would rather drop the linear option from both stoch_simul and model, since I still don't understand its purpose.
-- Sébastien Villemot Researcher in Economics & Debian Maintainer http://www.dynare.org/sebastien Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594
Thanks Johannes,
On 07/06/2012 09:47 AM, Johannes Pfeifer wrote:
- I agree with Sébastien to drop the option. From my experience it creates more confusion than it's worth. The confusion in this specific case might arise from the manual saying that one should rather put "linear" in model(linear). This suggests that it is indeed equivalent to stoch_simul(linear). However, the manual says that model(linear) sets order=1.
This seems to be the majority opinion so far...
- By the way: the manual states that using model(linear) spares one from specifying initval. This is not really true as linearity per se implies that there can be constant terms that need to be taken care of. In particular, I have seen a few mod-files in the forum that use a Taylor rule with the target interest rate as a constant. Hence, although the model is linear, the steady state of the interest rate is not 0 and Dynare crashed.
If the model is linear, Dynare computes the steady state directly using linear algebra, without needing initval, even if there are constants in the model. Could you please point me to an example where Dynare crashes in such a case? Thanks in advance.
Best
Michel
-- Johannes Pfeifer Haußerstr. 29 72076 Tübingen Tel.: +49-(0)7071-6396184 Mobil.: +49-(0)170-6936820 Germany
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Sébastien Villemot Gesendet: Donnerstag, 5. Juli 2012 18:15 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Michel Juillardmichel.juillard@mjui.fr writes:
Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
I would rather drop the linear option from both stoch_simul and model, since I still don't understand its purpose.
-- Sébastien Villemot Researcher in Economics& Debian Maintainer http://www.dynare.org/sebastien Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Dear Michel,
if you comment out the initval-block in the attached file, Dynare does not solve the model anymore. Best,
Johannes
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Michel Juillard Gesendet: Freitag, 6. Juli 2012 12:04 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Thanks Johannes,
On 07/06/2012 09:47 AM, Johannes Pfeifer wrote:
- I agree with Sébastien to drop the option. From my experience it creates more confusion than it's worth. The confusion in this specific case might arise from the manual saying that one should rather put "linear" in model(linear). This suggests that it is indeed equivalent to stoch_simul(linear). However, the manual says that model(linear) sets order=1.
This seems to be the majority opinion so far...
- By the way: the manual states that using model(linear) spares one from specifying initval. This is not really true as linearity per se implies that there can be constant terms that need to be taken care of. In particular, I have seen a few mod-files in the forum that use a Taylor rule with the target interest rate as a constant. Hence, although the model is linear, the steady state of the interest rate is not 0 and Dynare crashed.
If the model is linear, Dynare computes the steady state directly using linear algebra, without needing initval, even if there are constants in the model. Could you please point me to an example where Dynare crashes in such a case? Thanks in advance.
Best
Michel
-- Johannes Pfeifer Haußerstr. 29 72076 Tübingen Tel.: +49-(0)7071-6396184 Mobil.: +49-(0)170-6936820 Germany
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Sébastien Villemot Gesendet: Donnerstag, 5. Juli 2012 18:15 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Michel Juillardmichel.juillard@mjui.fr writes:
Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
I would rather drop the linear option from both stoch_simul and model, since I still don't understand its purpose.
-- Sébastien Villemot Researcher in Economics& Debian Maintainer http://www.dynare.org/sebastien Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Thanks Johannes,
in this model the problem comes from the 2 unit roots. It doesn' t make sense to ask Dynare to solve for the steady state of a non stationary model.
With initval, the user selects one of an infinite number of solutions for the static equations. But even with initval, it would not work if initval did not provide the exact solution for obs_r_h.
I will add at some point a discussion of models with non-stationary variables, but probably not in the paragraph about linear option of model keyword
All the best,
Michel
On 07/06/2012 12:21 PM, Johannes Pfeifer wrote:
Dear Michel,
if you comment out the initval-block in the attached file, Dynare does not solve the model anymore. Best,
Johannes
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Michel Juillard Gesendet: Freitag, 6. Juli 2012 12:04 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Thanks Johannes,
On 07/06/2012 09:47 AM, Johannes Pfeifer wrote:
- I agree with Sébastien to drop the option. From my experience it creates more confusion than it's worth. The confusion in this specific case might arise from the manual saying that one should rather put "linear" in model(linear). This suggests that it is indeed equivalent to stoch_simul(linear). However, the manual says that model(linear) sets order=1.
This seems to be the majority opinion so far...
- By the way: the manual states that using model(linear) spares one from specifying initval. This is not really true as linearity per se implies that there can be constant terms that need to be taken care of. In particular, I have seen a few mod-files in the forum that use a Taylor rule with the target interest rate as a constant. Hence, although the model is linear, the steady state of the interest rate is not 0 and Dynare crashed.
If the model is linear, Dynare computes the steady state directly using linear algebra, without needing initval, even if there are constants in the model. Could you please point me to an example where Dynare crashes in such a case? Thanks in advance.
Best
Michel
-- Johannes Pfeifer Haußerstr. 29 72076 Tübingen Tel.: +49-(0)7071-6396184 Mobil.: +49-(0)170-6936820 Germany
-----Ursprüngliche Nachricht----- Von: dev-bounces@dynare.org [mailto:dev-bounces@dynare.org] Im Auftrag von Sébastien Villemot Gesendet: Donnerstag, 5. Juli 2012 18:15 An: List for Dynare developers Betreff: Re: [DynareDev] linear option in stoch_simul
Michel Juillardmichel.juillard@mjui.fr writes:
Fujiwara, Ghironi and Bilbiie sent me a message saying that stoch_simul(linear) doesn't report the same results in 4.3 as in the previous version. But they use it with a nonlinear model! I check the manual and we say explicitly that LINEAR declares that the model is linear not to use a linear approximation. I guess that before version 4.3.0, we forced order=1 when the linear option was used in stoch_simul. But, as it is both contrary to the manual and confusing, I would be tempted of forbidding LINEAR as an option of stoch_simul, but this will break existing *.mod files. Or we can be nice to users, and keep interpreting LINEAR as identical to order=1, but we must correct the manual.
What do you think? We may need to check older documentation to understand how the confusion got about.
I would rather drop the linear option from both stoch_simul and model, since I still don't understand its purpose.
-- Sébastien Villemot Researcher in Economics& Debian Maintainer http://www.dynare.org/sebastien Phone: +33-1-40-77-84-04 - GPG Key: 4096R/381A7594
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
-
Johannes Pfeifer -
Michel Juillard -
Sébastien Villemot