[Gretl-users] Exact method of likelyhood

Good afternoon. I am developing a causal econometric model according to the concept of аuthenticity modeling Model is as follows: Y(t)=SUM[a0*Y(t-s)] + SUM[a6*X6,(t-s)] + SUM[a23*X23,(t-s)] (1->4) (0->2) (0->1) The right side of this equation has lag dependent varibles, lag independent variables and independent variables. To estimate parameters I use the exact method of likelyhood. But the residuals are not normally distributed. May I use this method to estimate the parameters (or I should use another methods)? If the residuals are not normal, will be this model suitable for forecasting? Thanks in advance Alexander