Am 25.03.2021 um 01:47 schrieb Reynaldo Senra: Right, I was able to take a look in the meantime. OK, interesting observation that most papers still report the short-run MG results. Although of course my guess is that they do that because easy implementations like Eviews print them out and people are simply afraid to _not_ report them, without any deep reason. Well, to sum up the discussion between Jack and me, for now you should get the desired results by doing: meanr(b.fullcoeffs) (where the "b" name is arbitrary and determined by you as the script author) And for the standard errors try this: sdc(b.fullcoeffs', b.N * (b.N - 1))' Some zeros will appear there in those places where the homogeneous long-run coefficients are given. Ignore those. Hope this works Sven