Hi, everyone,

I'm trying to perform Engle-Granger 2-step procedure. However, I did not find an appropriate article for that purpose and so, I would like to ask you about that. The STATA software has an option to perform Engle-Granger and ECM ( egranger var names, reg and egranger var name, ecm reg) s this right? I would also need your advice about how to use this option properly. First I have to perform unit root tests i.e. dickey-fuller and Philips perron tests.
Suppose that I find that my variables are non-stationary at levels, but using the first difference they are I(1). Then I have to perform Engle-Granger test. In case that Philips Perron suggests me 2 lags, I have to perform the test by using 2 lags or not? And I have also another question the first-step Engle-Granger must show me that D.e- gresid is negative and significant in the second-step equation, which is the ECM in order to have significant results the D-egresid must again be negative and statistical significant? I mean if it is positive and statistical significant the ECM is not right, even if it gives me good results for the examined variables? And finally, the first-step regression shows the long-run effects and the second step the short-run effects or I misunderstood something?


I'm sorry for my long extensive question, but I'm really confused.