Trends and dummies in IRIS

[pfjulio]

Dear Jaromir,

Suppose we have a backward looking semi-structural model featuring several equations with error correction mechanisms implemented in IRIS.

Can you include trends and/or dummies in the cointegrating relationships?

How do you implement this?

Kind regards

[jaromir-benes]

Hi

you can specify the trends and dummies as some mechanical processes (or even use shocks for dummies), e.g.

tnd = tnd{-1} + c + shk_tnd; dummy = 1 + shk_dummy;

and then use these variables in your other equations.

Both for simulation and estimation, you can then tune the paths of these processes to desired values.

Does this help? Or do you need something else?

Best J

On Tue, Aug 30, 2022 at 4:44 PM pfjulio @.***> wrote:

Dear Jaromir,

Suppose we have a backward looking semi-structural model featuring several equations with error correction mechanisms implemented in IRIS.

Can you include trends and/or dummies in the cointegrating relationships?

How do you implement this?

Kind regards

— Reply to this email directly, view it on GitHub https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGCVKKXDSA7FLXWVXMWK43TV3YM4NANCNFSM6AAAAAAQAPLONQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[pfjulio]

Yes that helps. One final question.

How do I tune the paths for those variables? Shall I treat them as exogenous processes and give them as observables? Or must I define some system property (in estimation)?

Many thanks

A quarta, 31/08/2022, 08:13, Jaromír Beneš @.***> escreveu:

Hi

you can specify the trends and dummies as some mechanical processes (or even use shocks for dummies), e.g.

tnd = tnd{-1} + c + shk_tnd; dummy = 1 + shk_dummy;

and then use these variables in your other equations.

Both for simulation and estimation, you can then tune the paths of these processes to desired values.

Does this help? Or do you need something else?

Best J

On Tue, Aug 30, 2022 at 4:44 PM pfjulio @.***> wrote:

Dear Jaromir,

Suppose we have a backward looking semi-structural model featuring several equations with error correction mechanisms implemented in IRIS.

Can you include trends and/or dummies in the cointegrating relationships?

How do you implement this?

Kind regards

— Reply to this email directly, view it on GitHub https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AGCVKKXDSA7FLXWVXMWK43TV3YM4NANCNFSM6AAAAAAQAPLONQ

. You are receiving this because you are subscribed to this thread.Message ID: @.***>

— Reply to this email directly, view it on GitHub https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334#issuecomment-1232551151, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE2OWEVAB23C6XGPOBVU6UTV34AYRANCNFSM6AAAAAAQAPLONQ . You are receiving this because you authored the thread.Message ID: @.***>

[jaromir-benes]

In estimation/kalman filtering, yes, you simply supply observables for the "exogenous" variables.

To get a proper behavior of the likelihood function, use the following two tricks:

1. Use the option "excludeFromObjFunc" (available right now only in the latest commits on the bleeding and stable branches, not in the official release yet) to list all the "exogenous" variables - you don't want the "goodness of fit" of their mechanical processes to influence the likelihood function.

2. Define the "exogenous" variables (the trends and dummies) so that they appear in the other equations only with a lag, not contemporaneously. This is no loss in generality because you simply move the variables by one period, including its observations. This is to make sure that when evaluating the likelihood function for the other variables at time t (which is based on a one-step ahead prediction), you use the correct values for the "exogenous" variables and not their predictions based on the mechanical processes.

Also, if you only want to discuss certain questions, please use rather the Discussion section instead of the Issues section (no big deal, it's just more convenient for everyone).

Jaromir

On Wed, Aug 31, 2022 at 1:48 PM pfjulio @.***> wrote:

Yes that helps. One final question.

How do I tune the paths for those variables? Shall I treat them as exogenous processes and give them as observables? Or must I define some system property (in estimation)?

Many thanks

A quarta, 31/08/2022, 08:13, Jaromír Beneš @.***> escreveu:

Hi

you can specify the trends and dummies as some mechanical processes (or even use shocks for dummies), e.g.

tnd = tnd{-1} + c + shk_tnd; dummy = 1 + shk_dummy;

and then use these variables in your other equations.

Both for simulation and estimation, you can then tune the paths of these processes to desired values.

Does this help? Or do you need something else?

Best J

On Tue, Aug 30, 2022 at 4:44 PM pfjulio @.***> wrote:

Dear Jaromir,

Suppose we have a backward looking semi-structural model featuring several equations with error correction mechanisms implemented in IRIS.

Can you include trends and/or dummies in the cointegrating relationships?

How do you implement this?

Kind regards

— Reply to this email directly, view it on GitHub https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334, or unsubscribe <

https://github.com/notifications/unsubscribe-auth/AGCVKKXDSA7FLXWVXMWK43TV3YM4NANCNFSM6AAAAAAQAPLONQ

. You are receiving this because you are subscribed to this thread.Message ID: @.***>

— Reply to this email directly, view it on GitHub < https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334#issuecomment-1232551151 , or unsubscribe < https://github.com/notifications/unsubscribe-auth/AE2OWEVAB23C6XGPOBVU6UTV34AYRANCNFSM6AAAAAAQAPLONQ

. You are receiving this because you authored the thread.Message ID: @.***>

— Reply to this email directly, view it on GitHub https://github.com/IRIS-Solutions-Team/IRIS-Toolbox/issues/334#issuecomment-1232831428, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGCVKKR4HRUB45LF3VLZ5QDV35BATANCNFSM6AAAAAAQAPLONQ . You are receiving this because you commented.Message ID: @.***>