On Sat, 31 Jan 2015, Stefan A. wrote: Unless I'm mistaken, this is quite easy; for example: <hansl> open denmark vecm 2 1 LRM LRY IBO IDE set horizon 40 cumIRF = cum($vma) </hansl> However, I have two econometric remarks on this (feel free to ignore them of course if they are not applicable to your case). 1 - you should be aware that the $vma matrix is just a mechanic inversion of the VAR parameters and has no meaningful economic interpretation per se. Apart from the problem of fundamentalness (Lippi-Reichlin), it is not even interpretable in terms of structural shocks, unless you make some sort of identification assumption, and the literature on this is enormous. If you wanted to replicate the "classic" IRF analysis you get by a recursive identification scheme, you'd have to do something like this: <hansl> open denmark vecm 2 1 LRM LRY IBO IDE set horizon 40 K = cholesky($sigma) IRF = $vma * (K ** I(4)) print IRF </hansl> 2 - I don't think you need to _cumulate_ the IRFs from a VECM: a VECM is just a way to rewrite a VAR that contains a few unit roots by contruction, which causes the IRFs to thave a non-zero limit without cumulating them. I fail to see any meaningful interpretation for a cumlated IRF (unless you're purposedly playing with I(2) trends). Hope this helps! ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------