Trying to figure out the projection/fitted equation for a GRETL ARMAX estimation with an AR(2) Lag (with complex roots) using the ML estimator An example below const 5.73988 0.0368359 155.8 0.0000 *** phi_1 0.439231 0.183336 2.396 0.0166 ** phi_2 −0.775841 0.147472 −5.261 1.43e-07 *** X1 0.665307 0.173553 3.833 0.0001 *** X2 −0.199019 0.0600511 −3.314 0.0009 *** X3 −0.151153 0.0637270 −2.372 0.0177 ** Does gretl apply the lag operator to the dependent variable and estimate Yt= c + p1Yt-1+p2Yt-2 +ΣbX +et or estimate Yt=c+ΣbX +ut in which Ut incorporates the AR2 structure (as in AR General Cochrane Orcutt estimation). In other words what is the final form of the equation to generate the fitted or future values(at least where past residuals would still apply) ?