Thanks, Prof. Sven.

I understand what you point out.

However, the kind of model I want to estimate is

y = a + b*x(x < mu) + c*x(x>=mu) + e

In the literature, the scalar mu is usually taken to have a threshold of zero. My aim is to determine mu through a grid search of x, let's say the exchange rate.

Thanks.

On Wed, 1 Feb 2023 at 16:28, Sven Schreiber <sven.schreiber@fu-berlin.de> wrote:

Am 31.01.2023 um 12:44 schrieb Riccardo (Jack) Lucchetti:
> On Tue, 31 Jan 2023, Olasehinde Timmy wrote:
>
>> Dear profs.,
>>
>> Please, I would like to know how to perform a grid search in order to
>> find
>> an optimal value for a threshold—for example, the threshold value to
>> choose
>> for the inflation rate in an asymmetric model.
>
> A grid search is relatively easy to perform via the "for" variant of
> the loop command. ...

The contributed function package SETAR.gfn by Federico Lampis & Ignacio
Díaz-Emparanza might also be useful and relevant, which implements stuff
from Hansen, "Sample Splitting and Threshold Estimation", Econometrica,
2000.

(BTW, in the help text of the current version 1.5 I think there are some
parentheses missing in the definition of the model:

Y_(t)= a1*Y_(t-1)+a2*Y_(t-2)+···+ap*Y_(t-p)(Y_(t-d)<=gamma) +
b1*Y_(t-1)+b2*Y_(t-2)+···+bp*Y_(t-p)(Y_(t-d)>gamma) + e_(t)

... but the meaning should be clear enough.)

cheers

sven
_______________________________________________
Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/