Panel Granger causality using dynamic error correction model

[jnow123]

Hi. I'm in trouble to estimate Panel GMM (Arellano and Bond, 1991).

I am trying to estimate Granger causality.

For this, I confirmed long-run relationship between variables and extracted ECT using FMOLS.

I wonder how to estimate this variables.

My questions are followed:

1. All variables are non-stationary so I have to put these in equation as differenced but I know ECT should not be differenced.

As I know, in dynamic wizard, there are two options when handling fixed effects. I think "difference" menu make all variables in equation as differenced form.

How can I maintain ECT as level form? Should I choose "orthogonal deviations"?

2. When I finished wizard and start estimation, Results show me the variables which I didn't add. (I didn't remember exactly but maybe these are the @lev"1994" form.

Sincerely.