On Fri, 2 Mar 2018, Marvin Berndt wrote: This should illustrate what gretl does: <hansl> open data9-7 list Y = INCOME UNEMP PRIME var 3 Y matrix U = $uhat # note: no df adjustment below V = U'U ./ $T eval det(V) logdet = log(det(V)) # number of equations g = nelem(Y) # coefficients per equation p = $ncoeff LN2PI = log(2*$pi) loglik = -(g * $T/2) * (LN2PI + 1) - ($T/2) * logdet # AIC calculated on per-observation basis AIC = (-2 * loglik + 2 * g * p) / $T </hansl> The manually calculated values agree with those printed by gretl. Allin Cottrell