[Gretl-devel] Re: Dirichlet distribution

On Sat, Jul 27, 2024 at 7:01 AM Riccardo (Jack) Lucchetti <p002264(a)staff.univpm.it&gt; wrote: > While browsing lib/src/random.c, I noticed that we've had the "dir" > parameter to mrandgen for a LOOOOOOOOONG time (if my reconstruction is > correct, since 2022a). For example > > <hansl> > matrix D = mrandgen("dir", {0.4, 1.2, 0.8}, 16) > </hansl> > > It would seem that we forgot to document our ability to draw from a > Dirichlet distribution; the Changelog is silent on this too. > > I can't find any email exchange on this subject. Does anyone remember if > this was a conscious choice (maybe, insufficient testing), or did we > just forget? I found the following message that I sent on Dec 16, 2021, which kinda suggests that it wasn't sufficiently tested: <quote> I've added Dirichlet ("dir") to the options for mrandgen(), but I'm not sure I've got this right; comments are welcome.