Dynamically Stable ARDL

[Amiro]

Hi,
I have an ARDL model(1,2,3). How can I test whether it is dynamically stable and the inverse roots are inside the unit circle?

Prof. David Giles suggested a trick to do so in Eviews a simple AR model. However, my model has independent variables with their own lag.

Is it correct to run the model after replacing all the lagged variables of ONLY the dependent variable with AR terms? So if in an unrestricted ECM with lags (1,2,3):

d(y)=B1+d(y(−1))+d(X1(−1))+d(X1(−2))+d(X2(−1))+d(X2(−2))+d(X2(−3))+Y(−1)+X1(−1)+X2(−1)+e(t)

(coefficients not added for simplicity)

So I only replace y(-1) with AR(1), and then check the AR structure?

2)

If I want to test the dynamic stability of a restricted ECM (where the level terms above are restricted), can I follow the same procedure to test for stability?

Thank you!!
A