So, I am trying to run a regression that requires me to log and take the first differences of it to render stationary. That is why I am using VECM.

I ran the diagnostic tests, and everything is fine except that I have heteroskedasticity and autocorrelation. So, I tried to use newey west SE in Stata and it fixed the problem. However, I have one question. Should I run let's say
newey d.log(y) d.log(x1) d.log(x2) lag() or only newey log(y) log(x1) log(x2) lag() to follow the VECM?

Thank you in advance!