Re: [Help-gsl] Question about the GSL random number generator-wishart (yorkey)

[Patrick Alken]

Hello,

  If you could provide files similar to the existing files in the 
randist directory for the two new distributions, I can certainly help 
get them added in.

I am no expert on random number generation so I probably can't help much 
with the testing. But a quick look at randist/test.c - it looks like the 
testMoments() generates N random numbers and counts how many fall within 
[a,b], and compares that with the expected number of p*N. The trick 
would be determining the appropriate a,b,p from the distribution you are 
using.

But also it looks like very few of the distributions use the 
testMoments() test. The testPDF() test looks more important.

Patrick

On 07/07/2016 10:21 AM, Timothée Flutre wrote:
> Hello,
>
> it would indeed be very useful to add support for random number generation
> for the multivariate Normal and Wishart distributions in the GSL.
>
> I recently obtained C code from colleagues who published an article using
> the GSL (Jia and Jannink, Genetics, 2012). It hence is under the GPL, too.
> In their code, they added 2 functions, one for multivariate Normal and one
> for Wishart.
>
> How can I add these 2 functions to the GSL? Can I simply create files ready
> to be included into the GSL and send them to this mailing list?
>
> Moreover, the original authors didn't provide any test. So I guess one
> would have to add this. So I had a quick look into "tree/randist/test.c" in
> the git repo, but didn't quite understand how "testMoments()" is working. I
> can't be considered as a solid C programmer... I can give it a try, but
> would anyone be willing to help?
>
> Timothée Flutre
>
> 2016-06-21 18:00 GMT+02:00 <address@hidden>:
>
> > Send Help-gsl mailing list submissions to
> >          address@hidden
> >
> > To subscribe or unsubscribe via the World Wide Web, visit
> >          https://lists.gnu.org/mailman/listinfo/help-gsl
> > or, via email, send a message with subject or body 'help' to
> >          address@hidden
> >
> > You can reach the person managing the list at
> >          address@hidden
> >
> > When replying, please edit your Subject line so it is more specific
> > than "Re: Contents of Help-gsl digest..."
> >
> >
> > Today's Topics:
> >
> >     1. Question about the GSL random number generator-wishart (yorkey)
> >
> >
> > ----------------------------------------------------------------------
> >
> > Message: 1
> > Date: Tue, 21 Jun 2016 10:38:33 +0800
> > From: yorkey <address@hidden>
> > To: address@hidden
> > Cc: yorkey <address@hidden>
> > Subject: [Help-gsl] Question about the GSL random number
> >          generator-wishart
> > Message-ID: <address@hidden>
> > Content-Type: text/plain; charset=gb2312
> >
> > Dear Sir or Madam,
> >       I am using the GSL random number generator to sample cases from
> > various distributions in my project. It is great to find that GSL supports
> > lots of distributions. However, neither ?wishart? nor ?inverse wishart? has
> > been found in GSL. In Gibbs sampling, we always draw samples from the
> > inverse wishart distribution, which is the conjugate prior for the
> > covariance matrix of the multivariate gaussian distribution. Notice that
> > the multivariate gaussian and the related conjugate prior distributions are
> > frequently employed in Bayesian analysis, I want to know: is there any
> > alternative implementation of wishart is GSL? Or is there any version under
> > development supporting the inverse wishart distribution?
> >      Thanks a lot for your kind attention.
> >
> > Best,
> > yorkey
> >
> >
> >
> > ------------------------------
> >
> > Subject: Digest Footer
> >
> > _______________________________________________
> > Help-gsl mailing list
> > address@hidden
> > https://lists.gnu.org/mailman/listinfo/help-gsl
> >
> >
> > ------------------------------
> >
> > End of Help-gsl Digest, Vol 151, Issue 2
> > ****************************************