#501 2D kernel density estimate with dgrid3d

[Péter Juhász]

I was trying to plot a kernel density estimate for a 2D dataset. This is possible for 1D data (with 'smooth kdensity'), but not for 2D data.

While trying to do it with external tools, I realized that the existing 'dgrid3d' functionality already has every ingredient, and in fact does exactly what I need with a small modification. (dgrid3d uses the following formula to calculate the new grid values: z = Sum_i w(d_i) * z_i / Sum_i w(d_i). If I set all z_i to a constant, and omit the division, I get a KDE.)

This patch implements a new option to 'set dgrid3d': 'kdensity2d'. If given, it modifies the algorithm so that the division mentioned above does not take place.

Documentation and new demo included.

[Ethan Merritt]

Looks good to me. Perhaps another sentence or two in the docs that defines or explains "kernel density"? I had to look it up separately.

[Péter Juhász]

Re-uploaded with slightly more documentation.
Feel free to reword it if it seems too un-English.

[Ethan Merritt]

Looks OK, except there is a missing word:
"this smooth [surface] is plotted instead of the original data"

It still remains to add the option to save.c and show.c

[Péter Juhász]

2D kernel density estimate with dgrid3d - v3

[Péter Juhász]

I've totally forgot about those, but it's fixed now.

[Ethan Merritt]

Are you going to go ahead and add this to CVS?

[Péter Juhász]

Thanks for the reminder, I've checked it it now.

[Nobody/Anonymous]

WZe7PA <a href="http://psnzghkjbqin.com/">psnzghkjbqin</a>, [url=http://hpspyawzyksh.com/]hpspyawzyksh[/url], [link=http://qreqknnbjqjo.com/]qreqknnbjqjo[/link], http://tjoearljdiuo.com/

[Douglas]

I could really use this feature. Is it in the CVS version? (And therefore, do I need to check it out) or has it been integrated into any official version?

[Péter Juhász]

It's in the CVS, but not in any released version.