Re: [Gretl-users] impulse-response functions from VECM in Grtel

Am 17.08.2017 um 13:27 schrieb lasses skola: Hi, I'm assuming that you are using the SVAR add-on, and that you are using the scripting interface, otherwise I don't see how you could get bootstrapping and VEC models together. (Disclosure: I'm the second author of the SVAR add-on, but the bulk of that and also of the documentation was written by Jack.) So, quoting from the documentation pdf file (section 2.4): "A word on the unit of measurement of IRFs: by their definition (see equation (7)), and the fact the structural shocks are assumed to have unit variance, clearly their unit of measurement is the same as the one for the corresponding observable variable y i,t." If you have a levels variable in billion dollars, the y-axis unit will be billion dollars. If you have a growth rate in percentages, so will be that y-axis. Perhaps your question is about the underlying shock definition. The IRF graphs here should be giving you the response of the respective variable to a unit structural shock. (People, correct me if I'm wrong, it's been a while since I looked at that add-on.) Sometimes people want to assume a different structural shock size, e.g. the size of one standard deviation of the i-th endogenous variable. This could be achieved by multiplying the IRFs with that SD number. (The doc also mentions a different normalization which really changes the y-axis units, if you're interested.) HTH, Sven