SPUG: RFC: Statistics::KernelEstimation - Kernel Density Estimates and Histograms

[Philipp K. Janert]

I would like to invite comments on a new module, named
Statistics::KernelEstimation.


This modules calculates Kernel Density Estimates and related
quantities for a collection of random points.

A Kernel Density Estimate (KDE) is similar to a histogram, 
but improves on two known problems of histograms: it is 
smooth (whereas a histogram is ragged) and does not suffer 
from ambiguity in regards to the placement of bins.

In a KDE, a smooth, strongly peaked function is placed at the 
location of each point in the collection, and the contributions 
from all points is summed. The resulting function is a smooth
approximation to the probability density from which the set of 
points was drawn.

This module calculates KDEs as well as Cumulative Density 
Functions (CDF). Three different kernels are available 
(Gaussian, Box, Epanechnikov). 

The module also includes limited support for bandwidth optimization.

Finally, the module can generate "classical" histograms and
distribution functions.
 

The full POD is available here:
        http://www.beyondcode.org/projects/kernelestimation/


Let me know what you think!

Best,

		Ph.