Information Technology Reference
In-Depth Information
0.2
AtMr
0.04
RAC
0.025
0
0.002
CmRb
0.2
0.03
0.02
0.1
0.015
0.08
Gaut
0.004
0.02
0.15
0.07
0.02
InAs
CrJk
0.15
DrgR
KZN
DrgR
CmRb
0.03
0.01
0.06
0.006
0.3
AtMr
0.1
InAs
PubV
CmAs
0.008
0.2
WCpe
0.1
Arsn
0.1
Mrd
0.006
CmAs
0.20.01
BRs
0
Rape
0.2
0.004
0.001
0.05
BNRs
0.06
0.002
0.015
Mpml
0.008
NWst
FrSt
0.04
0.04
0.22
0.05
0.25
ECpe
0
Limp
0.01
0.0015
0.01
0.3
NCpe
0
0.05
0.02
0.012
0.35
AGBH
PubV
Figure 7.20
Two-dimensional CA biplot of 2007/08 crime data set. Approximating the
row profiles
R
−
1
(
X
-
E
) by plotting
R
−
1
/
2
U
and
C
1
/
2
V
(case B) with calibrations
scaled to provide predictions of actual profile values by setting the argument
Row-
Prof.scaled.markers = TRUE
in the function
cabipl
. All other arguments are kept
at their Figure 7.19 values.
The reader can check with our function
ca.predictivities
that
Gaut
,
KZN
and
WCpe
have row predictivities of 0.9735, 0.5366 and 0.9499, respectively. The three-
dimensional row predictivity of
KZN
is 0.8479. The reader can also verify that the
two-dimensional axis predictivities of
DrgR
and
Mrd
are 0.8871 and 0.1482, respectively.
If a third dimension is added the three-dimensional column predictivity of
Mrd
increases
sharply to 0.8469. This suggests that looking at a biplot with the second and third
eigenvectors as scaffolding might be worthwhile. So changing, in the argument list of
cabipl
,
e.vects = 2:3
, we obtain the CA biplot given in Figure 7.29.