Information Technology Reference
In-Depth Information
SE
IRD
-0.6
-3.5
-1
-0.5
-0.4
-2.5
−
4
3
-0.15
14
16
-3.5
-2
-3
4
-0.4
-2.5
5
2
1
-2
-1.5
-1.5
15
10
9
-1
1
7
18
12
-0.5
8
7
11
19
0
13
-0.4
-0.2
-0.3
-0.5
0.5
EDM
20
6
0
-1.5
0.5
0.5
-0.2
1
-0.2
EDSA
ALCO
CM
MM
Figure 3.1
PCA biplot of 95% VAR of financial instruments data constituting a
portfolio.
called with very few changes to the default settings; then the biplot is fine-tuned by
utilizing the many arguments available for that purpose. As an example, consider the
number of scale markers on each biplot axis in Figure 3.1. Some of the axes, like
MM,
have too few markers while others, like
EDM,
perhaps have too many. The argument
n.int
provides a mechanism for equipping each axis with just the right number of
markers. This has been done in Figure 3.2.
The PCA biplot in Figure 3.1 is a useful multidimensional scatterplot represent-
ing the seven financial instruments over 20 days in a single display. Since the biplot
axes shown here are prediction biplot axes, the approximate values for the individual
financial instruments can be readily read off from the axes. Figure 3.2 illustrates how
predictions for selected sample points can be obtained with
PCAbipl
using arguments
predictions.sample
and
ort.lty
. The following changes have been made in the call
to
PCAbipl
to obtain Figure 3.2:
n.int = c(3,10,10,5,10,10,3), predictions
.sample = c(2,16), ort.lty = 2.
By default, the specified predictions are shown on the biplot as illustrated in Figure 3.2
and also returned as output on the R console - see Table 3.2. We note that on
day2
(and
also on
day3
)abs(VAR)of
EDM
was particularly large; this is in sharp contrast with
