Thanks very much for the feedback.
Best regards
    On Monday, March 17, 2025 at 11:45:29 PM EDT, Sven Schreiber 
<sven.schrei...@fu-berlin.de> wrote:  
 
 Am 17.03.2025 um 09:11 schrieb Riccardo (Jack) Lucchetti:
> On 17/03/2025 00:01, Olivo Victor wrote:
>> Hello Everyone:
>>
>> Is there a simple way to get and save the residuals in the Bai-Perron 
>> and thresh_infer 0.1 packages?
>
> For ther Bai-Perron package,

Ahem, for the record, Jack, our package is called "StrucBreak".

About thresh_infer: Thanks for the feedback; neither the basic inputs 
like the specified regressor list nor auxiliary output like the 
residuals are currently stored in the returned bundle, and they probably 
should be in the next version of the package.

Currently, I think you can get the residuals for the estimated 
two-regime model as follows:

= y - (b.whichregime==1) * lincomb(const || rhs, b.Regime1.coeff) - 
(b.whichregime==2) * lincomb(const || rhs, b.Regime2.coeff)

Here, "y" stands for the name of your dependent variable series, "rhs" 
stands for the list of regressors, and "b" is your chosen name for the 
returned bundle from the H_thresh_estim call. Adjust the line 
accordingly. All the other names and identifiers are given by the 
thresh_infer package. (Except "lincomb", which is a gretl function for 
doing --guess what-- a linear combination.) OK, the line is a bit long, 
but quite intuitive and self-explanatory, I think. Please verify if the 
result makes sense, because I haven't really tested it yet.

cheers

sven

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

Reply via email to