EViews.com

Dear Gareth,

EViews offers the capability to estimate an ARDL model by optimizing the number of lags based on criteria like AIC, BIC, etc. It also allows viewing the ranking of different specifications (Model Selection Criteria Table). Is it possible to program one's own specification of lags to incorporate our expertise, which might not be fully taken into account by automated model selection criteria? For instance, if
equation toto.ardl(deplags=12, reglags=12, ic=aic) y x1 x2
suggests an ARDL(3,3,3) but I want an ARDL(3,1,1). Of course, I should be able to manage this with LS, but then I would lose access to other features like toto.boundstest...

Thank you in advance for your time and valuable assistance.
Best regards,
Bob

Use fixed lags.

I tried but I can't manage to do an ARDL(3,1,2), only an ARDL(3,1,1) or (3,2,2)... with:
toto.ardl(fixed, deplags=3, reglags=1) y x1 x2
toto.ardl(fixed, deplags=3, reglags=2) y x1 x2
?

Ah, you want separate lags for different regressors. Yeah, no :(

Thank you!