search

jmaih / RISE_toolbox

Solution and estimation of Markov Switching Rational Expectations / DSGE Models
BSD 3-Clause "New" or "Revised" License
105 stars 77 forks source link

check plot #148

Closed drepalma closed 3 years ago

drepalma commented 4 years ago

Dear Junior,

I would like to know if RISE has any options similar to the check plot available in Dynare?

Thank you in advance for your attention! best regards from Brazil Andreza

jmaih commented 4 years ago

Dear Andreza,

In RISE the function is called mode_curvature. For details, check the help by running

help dsge.mode_curvature

Cheers,

J.

On Fri, Sep 4, 2020 at 3:11 PM drepalma notifications@github.com wrote:

Dear Junior,

I would like to know if RISE has any options similar to the check plot available in Dynare?

Thank you in advance for your attention! best regards from Brazil Andreza

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/jmaih/RISE_toolbox/issues/148, or unsubscribe https://github.com/notifications/unsubscribe-auth/AATKBTYNUEUMKFEMCTL5BL3SEDROZANCNFSM4QYKN5UQ .

drepalma commented 4 years ago

Dear Junior,

Thank you very much! This worked, but the results that I found are not very adequate. How can I improve the mode_curvature? Should I change the priors or change the optimizer? Could you give me some guide? Thank you very much in advance for your help and attention. Best Regards , andreza

jmaih commented 4 years ago

Hi Andreza,

The mode curvature does what it does independently of the optimization procedure used. If for some reason the optimization does not reach a peak, this will most likely show up in your mode curvature.

Your intuition is correct that changing priors to more informative ones may help... by masking the deficiencies of the optimization procedure.

If you don't have a better optimizer, that might be one way to go.

The most important question, however, is not about whether posterior maximization was successful or not. The real question is what you plan to do with your estimates. For instance, if your ultimate goal is to do posterior simulation, you don't need to reach a peak, you don't even need to do posterior maximization first : the peak you are looking for in the posterior maximization phase is just one point in the entire posterior surface.

But if having nice-looking pictures of the mode curvature is a goal in itself, well, you need a good optimization procedure and good priors.

Cheers,

J.

On Mon, Sep 7, 2020 at 6:51 PM drepalma notifications@github.com wrote:

Dear Junior,

Thank you very much! This worked, but the results that I found are not very adequate. How can I improve the mode_curvature? Should I change the priors or change the optimizer? Could you give me some guide? Thank you very much in advance for your help and attention. Best Regards , andreza

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/jmaih/RISE_toolbox/issues/148#issuecomment-688434482, or unsubscribe https://github.com/notifications/unsubscribe-auth/AATKBT4HVTT6VUULVYFBEW3SEUFQPANCNFSM4QYKN5UQ .