EViews.com

Does anyone know how to minimize AIC or BIC when constructing an ARDL (Autoregressive Distributed Lag) model in Eviews?
I have seen similar questions posed in other posts in this forum. Unlike those posts I would like to try let's say 10 or 11 explanatory variables in the regression. Maximum lag length could be 8 quarters. I don't expect all variables to fit in the final model but it could be worth trying all of them.
Thankful for answers.

If there is anyone who has an answer to this question, I would be really thankful. I'm in urgent need!

Dear Nils,

Since you are urgently need the code for any purposes, I suggest you to use the following procedure to determine the ARDL model.

Assumed that you have 2 variables.

Step 1: Run an autoregressive process (AR) regression that is Y regress with the past value (t-j) of itself (Y). Then use the AIC or BIC to determine the best lag structure.

Step 2: Run a second regression with the Y and the selected lag length (fix lags for Y), then added the lagged X into the testing equation. By doing that you can determine the optimal lag order for your X with AIC and BIC.

The above approach is similar to the Hsiao's (1981) non-parametric causality test - Journal of Monetary Economics

Good luck and hope the above procedure helped you to solve your urgent problem.

Thank you,

Regards,
CF Tang

Thank you very much for your help!