I have run the Granger Causality in Gretl using VAR, after that I found there are unit roots in variables. So I changed to VECM model to test the causality between variables for example in following equation: Y =X1+X2+X3. I input Y as endogenous variable and X1 X2 X3 as exogeous variables. But the results didn't come up, it shows ˇ°You must select two or more endogenous variablesˇ± But in the original regression equation only Y is endogenous variable. why it shows that?

Does it make any difference for the results of Granger Causality if there are unit roots in variables? what should I avoid them instead of  using VECM in Gretl? 

Thank you for your time 

Dianpeng YU