KDE: Lisp-Stat objects for Kernel Density Estimation

Description

KDE means Kernel Density Estimation. We have developed a set of Lisp-Stat objects and functions for doing Kernel Density Estimation in a truly interactive way.

KDE can be used interactively with a standard mouse/keyboard interface or can be run from another Lisp-Stat program.

When used interactively, a KDE object creates a window to display estimates and other information. A typical KDE window look like this:

Among other options, the user can decide to display (or not to display) in a KDE window, together with a summary of the estimation being shown,

The kernel estimate of the density for the given data set.
The histogram (or frequency polygon). Number of bins to use is automatically computed, but it is also posible to interactively change it.
The data points in the upper border of the graphic.
A small plot of the kernel function in use.

Some animations made using KDE

They are for the Marron-Wand (JASA 92 paper) distributions number 3,4,6,7,8,9. They depict the density function and kernel density estimators obtained from a random saple of size N=300.

How to get it

You may get the most recent version (it may be quite old) of the software here.

A paper describing KDE objects in more detail can be dowloaded in postscript format, but it's a bit outdated.

How to run it

To run KDE you need Lisp-Stat , a Lisp dialect oriented to statistics and dynamic graphics.

It can be run on several platforms including Macintosh, MS-Windows, X-Windows/Unix, etc.

If you have Lisp-Stat and get the KDE files, you'll find a README file with detailed instructions for running all this.

MS-Windows users: If you don't have Xlisp-Stat installed in your computer, and your system is MS-Windows(32bits) like Windows95 or newer, the easiest way to get KDE running on it is to download the file cpkde.zip (http://libiya.upf.es/gaussFtp/stat/kde/cpkde.zip), to unzip it in a fresh directory (preserving directory structure) and double-click on wxls32.exe. Then use the 'Kernel Smooth' menu and the menu owned by each of the KDE windows you'll see.

Author

Frederic Udina contact me (sending mail to frederic dot udina at upf.edu.)