Page 162 - Computational Statistics Handbook with MATLAB

Chapter 5: Exploratory Data Analysis 149

The results are shown in Figure 5.26. Several argument options are available
for the plotmatrix function. If the first two arguments are matrices, then
MATLAB plots one column versus the other column. In our example, we use
a single matrix argument, and MATLAB creates scatterplots of all possible
pairs of variables. Histograms of each variable or column are shown along
the diagonal of the scatterplot matrix. Optional output arguments allow one
to add a title or change the plot as shown in the following MATLAB com-
mands. Here we replace the histograms with text that identifies the variable
names and display the result in Figure 5.27.
% Create the labels as a cell array of strings.
labs = {'Sepal Length','Sepal Width',...
'Petal Length', 'Petal Width'};
[H,ax,bigax,P] = plotmatrix(virginica);
axes(bigax)
title('Virginica')
% Delete the histograms.
delete(P)
%Put the labels in - the positions might have
% to be adjusted depending on the text.
for i = 1:4
txtax = axes('Position',get(ax(i,i),'Position'),...
'units','normalized');
text(.1, .5,labs{i})
set(txtax,'xtick',[],'ytick',[],...
'xgrid','off','ygrid','off','box','on')
end

ssuurr fafa
rf
s
dd
Iso
cceess
Sl
SSll
Sli ii iccesanand dIso IsoIso suur afac cees s
es
es
ccesanan
If we have a function defined over a volume, f xy z,,( ) , then we can view it
using the MATLAB slice function or the isosurface function (available
in MATLAB 5.3 and higher). This situation could arise in cases where we
have a probability density function defined over a volume. The slice capa-
bility allows us to view the distribution of our data on slices through a vol-
ume. The isosurface function allows us to view 3-D contours through our
volume. These are illustrated in the following examples.
Example 5.18
(
,,
To illustrate the slice function, we need f xy z) values that are defined
over a 3-D grid or volume. We will use a trivariate normal distribution cen-
tered at the origin with covariance equal to the identity matrix. The following
MATLAB code displays slices through the x = 0 , y = 0 , and z = 0 planes,
and the resulting display is shown in Figure 5.28. A standard normal bivari-
© 2002 by Chapman & Hall/CRC