Re: Bug in geomean()

[Jordi Gutiérrez Hermoso]

On 1 October 2012 10:36, kyfsympsn <address@hidden> wrote:
> Hi,
>
> Think I've noticed a big in the geomean(x) function (which can be also
> written as mean(x,"g"))
>
> In matlab, the following code:
>   x = rand(1000,1);
>   y=geomean(x)
>   z=exp(mean(log(x)))
>
> ...will give the same value for y and z (roughly 0.3664)
>
> However in octave y =0!!

Geomean is implemented in the NaN package. It is not part of core
Octave. Should it be? Is it part of core Matlab?

> I think octave is probably calculating the geometric mean by

Think no more, know:

    
http://hg.savannah.gnu.org/hgweb/octave/file/e8b117b405f9/scripts/statistics/base/mean.m#l112

Or from Octave:

    edit mean
    type mean

I think it's a good habit to inspect the source code if you want to
know how something is being done.

Can you provide a patch? It should take into account the cases of
complex, negative, or zero entries.

- Jordi G. H.