Re: [Gretl-users] irf() bootstrap method

Am 25.10.2013 03:22, schrieb logan.j.kelly: > > > Hello, > > One more question about irf(). Is there anything special about the > bootstrapping method it uses? I have tried to replicate it with a > combination of varsimul() resample() and irf() without the optional > alpha, and the confidence bands I get are narrower than that irf() > with the alpha argument supplied. > As Jack mentioned on this (or the devel?) list, bootstrapping irfs isn't trivial, there's quite a bit of literature on that. So I'm not surprised your results are different (and probably wrong, strictly speaking :-) Nonetheless, it's of course a valid question how gretl's irf() does it and where it's documented (apart from the source).