Graphics Reference
In-Depth Information
Figure
.
.
Kernel density estimate of the Toxic Release Inventory (TRI) sites in the US state of
Louisiana. Dark regions:areasofhighdensity.hick black line: path of the Mississippi River
estimate from Colorado is displayed using the black contours while the density es-
timate for Kansas is displayed using the grey contours. To keep the plot from be-
coming overly complicated, a small number of specifically chosen contours is used.
In particular, three contours are computed for each density and these contours con-
tain roughly
,
, and
% of the data. Multimodality is clearly present in both
estimates, representing different regions of the two states. he much more compact
density of the Kansas data is strongly representative of its much greater geographic
homogeneity relative to Colorado.
Higher-dimensional Density Estimates
5.3
hree- and
-D kernel density estimates are quite feasible theoretically. In the mul-
tivariate setting,
x
R
d
,thekernel,K
(
t
)
, is a multivariate density function oten
chosen to be a product kernel, defined by
d
j
=
K
(
t
)=
K
(
t
j
)
(
.
)
