ARDL ECM Cointegration Bound testing approach Pesaran

[philipvandelinde]

Hi all,

I have a question regarding the ARDL ECM bound approach within eviews for a study I am currently doing. I would be great if you could help me out!

I have one dependent variable (Eurzone trade balance) and four dependent variables (EuroGDP, China GDP, Real exchange rate and ex. rate volatility).

I have already done the ADF test and found all the variables to be I(0) or (1), as needed.

Now I want to perform the bound test on the OLS regression of the unrestricted ARDL ECM model, but I have no idea how to determine the optimal amount of lags for each different variable within this model before doing the OLS. Does anyone know how to do this within eviews?!
I have already experimented with different lag structures and found cases where the variables were cointegrated. I want to be precise with the number of lags however.

After determining this cointegrated relationship, I need to construct the restricted ARDL ECM model where the ECMt-1 variable needs to be included.
In order to construct this ECMt-1 variable, do I need to estimate the long run (equilibrium) model (e.g.) y(t)=a+b1x1(t)+b2x2(t)...etc. and save the residual of this regression and include it in the restricted ARDL ECM model or do I need to use the residuals of the OLS on y(t-1)=a+b1x1(t-1)+b2x2(t-1)... since the ECM is also (t-1) in the new ARDL restricted ECM model..? Please help!

Next, we I have the right ECMt-1 variable, can I use the same number of lags as I did in the model when performing the bound test or do I need to determine a new/different number of lags per variable. And if so, how do I do this in eviews?

Please help me out! Would be great if you could help me with this! Thank you so much!

Regards,
Philip