Page 322 - Computational Statistics Handbook with MATLAB

Chapter 8: Probability Density Estimation 311

8.6 MATLAB Code

The MATLAB Statistics Toolbox does not have any functions for nonparamet-
ric density estimation. The functions it has for estimating distribution param-
eters (e.g., mle, normfit, expfit, betafit, etc.) can be used for
parametric density estimation. The standard MATLAB package has func-
tions for frequency histograms, as explained in Chapter 5.
We provide several functions for nonparametric density estimation with
the Computational Statistics Toolbox. These are listed in Table 8.4.

TA
T AB B LE E E 8.4 8.4
T
A
A
T
E
8.4
8.4
L
B
BL
L
List of Functions from Chapter 8 Included in the Computational Statistics
Toolbox
Purpose MATLAB Function
These provide a bivariate histogram. cshist2d
cshistden
This returns a frequency polygon density csfreqpoly
estimate.
This function returns the Averaged csash
Shifted Histogram.
These functions perform kernel density cskernnd
estimation. cskern2d
Create plots csdfplot
csplotuni
Functions for finite and adaptive csfinmix
mixtures csadpmix

8.7 Further Reading

The discussion of histograms, frequency polygons and averaged shifted his-
tograms presented in this book follows that of Scott [1992]. Scott’s book is an
excellent resource for univariate and multivariate density estimation, and it
describes many applications of the techniques. It includes a comprehensive
treatment of the underlying theory on selecting smoothing parameters, ana-

© 2002 by Chapman & Hall/CRC