Re: [DynareDev] [GPM Maintainers] Fwd: Release of Dynare 4.2.4
Hi Sebastien, Michel, It looks like there are still some things wrong in the new version. When we calculate the RMSEs (using FilteredVariablesKStepAhead), there are huge differences in some variables between 4.2.4 and an older version (1 quarter ahead RMSE for BLT_EU comes out to be 400 instead of 14). Could you please check? Thank you and Best, Kadir -----Original Message----- From: gpm-maintainers-bounces@cepremap.org [mailto:gpm-maintainers-bounces@cepremap.org] On Behalf Of Sébastien Villemot Sent: Friday, December 02, 2011 10:31 AM To: GPM Maintainers Subject: [GPM Maintainers] Fwd: [DynareDev] Release of Dynare 4.2.4 Hi, We made a new release of Dynare which fixes a critical bug in 2nd order approximation, and a few other minor things. You should really not use 4.2.3, so we strongly encourage you to upgrade to 4.2.4. For estimating GPM6, you may still encounter issues with inverse gamma distribution priors with small variances. Previously Dynare was sometimes reporting wrong results in these cases: it is the new behavior which is correct. If you get an error, you should either increase the prior variance or calibrate the parameter. Best,
Hi Kadir, I will look into it. Best Michel On 12/8/2011 10:11 PM, Tanyeri, Kadir wrote:
Hi Sebastien, Michel,
It looks like there are still some things wrong in the new version. When we calculate the RMSEs (using FilteredVariablesKStepAhead), there are huge differences in some variables between 4.2.4 and an older version (1 quarter ahead RMSE for BLT_EU comes out to be 400 instead of 14).
Could you please check?
Thank you and Best, Kadir
-----Original Message----- From: gpm-maintainers-bounces@cepremap.org [mailto:gpm-maintainers-bounces@cepremap.org] On Behalf Of Sébastien Villemot Sent: Friday, December 02, 2011 10:31 AM To: GPM Maintainers Subject: [GPM Maintainers] Fwd: [DynareDev] Release of Dynare 4.2.4
Hi,
We made a new release of Dynare which fixes a critical bug in 2nd order approximation, and a few other minor things.
You should really not use 4.2.3, so we strongly encourage you to upgrade to 4.2.4.
For estimating GPM6, you may still encounter issues with inverse gamma distribution priors with small variances. Previously Dynare was sometimes reporting wrong results in these cases: it is the new behavior which is correct. If you get an error, you should either increase the prior variance or calibrate the parameter.
Best,
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Hi Kadir, I can't reproduce the problem. I attach my driver file and 2 smoother reports. One with original (?) May19 estimated parameters and one with parameters re-estimated on Nov21. For BLT_EU I obtain RMSE in t+1 around 17 - 26, but not 400. If you can reproduce the problem that you encountered, the best would be for you to run it on gpm.cepremap.org, so that we can look directly at the files. Best Michel On 12/08/2011 10:11 PM, Tanyeri, Kadir wrote:
Hi Sebastien, Michel,
It looks like there are still some things wrong in the new version. When we calculate the RMSEs (using FilteredVariablesKStepAhead), there are huge differences in some variables between 4.2.4 and an older version (1 quarter ahead RMSE for BLT_EU comes out to be 400 instead of 14).
Could you please check?
Thank you and Best, Kadir
-----Original Message----- From: gpm-maintainers-bounces@cepremap.org [mailto:gpm-maintainers-bounces@cepremap.org] On Behalf Of Sébastien Villemot Sent: Friday, December 02, 2011 10:31 AM To: GPM Maintainers Subject: [GPM Maintainers] Fwd: [DynareDev] Release of Dynare 4.2.4
Hi,
We made a new release of Dynare which fixes a critical bug in 2nd order approximation, and a few other minor things.
You should really not use 4.2.3, so we strongly encourage you to upgrade to 4.2.4.
For estimating GPM6, you may still encounter issues with inverse gamma distribution priors with small variances. Previously Dynare was sometimes reporting wrong results in these cases: it is the new behavior which is correct. If you get an error, you should either increase the prior variance or calibrate the parameter.
Best,
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Thank you very much, Michel. I will try it there and let you know... Best, Kadir -----Original Message----- From: Michel Juillard [mailto:michel.juillard@mjui.fr] Sent: Sunday, December 11, 2011 12:18 PM To: List for Dynare developers Cc: Tanyeri, Kadir; Sébastien Villemot; Garcia-Saltos, Roberto; Carabenciov, Ioan Subject: Re: [DynareDev] [GPM Maintainers] Fwd: Release of Dynare 4.2.4 Hi Kadir, I can't reproduce the problem. I attach my driver file and 2 smoother reports. One with original (?) May19 estimated parameters and one with parameters re-estimated on Nov21. For BLT_EU I obtain RMSE in t+1 around 17 - 26, but not 400. If you can reproduce the problem that you encountered, the best would be for you to run it on gpm.cepremap.org, so that we can look directly at the files. Best Michel On 12/08/2011 10:11 PM, Tanyeri, Kadir wrote:
Hi Sebastien, Michel,
It looks like there are still some things wrong in the new version. When we calculate the RMSEs (using FilteredVariablesKStepAhead), there are huge differences in some variables between 4.2.4 and an older version (1 quarter ahead RMSE for BLT_EU comes out to be 400 instead of 14).
Could you please check?
Thank you and Best, Kadir
-----Original Message----- From: gpm-maintainers-bounces@cepremap.org [mailto:gpm-maintainers-bounces@cepremap.org] On Behalf Of Sébastien Villemot Sent: Friday, December 02, 2011 10:31 AM To: GPM Maintainers Subject: [GPM Maintainers] Fwd: [DynareDev] Release of Dynare 4.2.4
Hi,
We made a new release of Dynare which fixes a critical bug in 2nd order approximation, and a few other minor things.
You should really not use 4.2.3, so we strongly encourage you to upgrade to 4.2.4.
For estimating GPM6, you may still encounter issues with inverse gamma distribution priors with small variances. Previously Dynare was sometimes reporting wrong results in these cases: it is the new behavior which is correct. If you get an error, you should either increase the prior variance or calibrate the parameter.
Best,
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
participants (2)
-
Michel Juillard -
Tanyeri, Kadir