Re: kernel density smoother?

[Søren Hauberg]

fre, 12 03 2010 kl. 13:51 +0000, skrev forkandwait:
> Søren Hauberg <soren <at> hauberg.org> writes:
> > That being said, I am curious: what would you like such a "dictator" to
> > do?
> 
> One of the best things about Python the language is the iron fist / velvet 
> glove with which "the standard library" is maintained and standardized. There 
> is very little need to go look for routines (like kernel density smoother), 
> since by the time a small brain like mine thinks of them, folks probably 
> figured out the package and included a standard way of doing it.  Octave 
> forge 
> plus File Exchange serves the role of putting everything in one place, but 
> not 
> of making it a consistent well documented bundle.  The "PEP" process is how 
> these things get figured out together.

I still don't quite understand the role of a "dictator". I agree, that
finding the right function for Octave can be a bit of a challenge from
time to time (especially if you are new to the tool; something us old
farts tend to forget from time to time). I also agree that documentation
for packages can be lacking. But I don't see how strong leadership could
change this (I guess this is a good indication that people shouldn't
declare me the Octave-Forge leader). Concrete ideas on how to improve
the situation are, however, most welcome as it is a problem that
concerns me.

> I think, esp if I don't know off 
> the top of my head that a kernel density smoother can be knocked together out 
> of a convolution.... ;)

Mathematically, a kernel density estimator is basically a convolution.
>From a programming point of view, things are a bit different as the
convolution functions assume equidistant samples, whereas a kernel
density estimator should work with arbitrary sampling schemes.

Søren