Information Technology Reference
In-Depth Information
0.1
CrJk
Gaut
0.3
Gaut
0.2
RAC
0.05
0.1
KZN
DrgR
WCpe
AtMr
InAs
CmRb
CmAs
BRs
0.3
WCpe
0.2
0.1
0
Mrd
Mpml
NCpe
NWst
FrSt
BNRs
Limp
Arsn
0.02
Rape
PubV
ECpe
AGBH
0.02
0.05
0.02
0.02
Mpml
0.04
Figure 7.21
Two-dimensional CA biplot of the 2007/08 crime data set but with the
transpose of the data matrix used as input. Approximating the row profiles
C
−
1
(
X
-
E
)
by
plotting
R
−
1
/
2
U
2
(case A) of the
transposed matrix
X
with argument
lambda = TRUE
(
λ
=
368
.
29, indicating that setting
lambda = FALSE
would result in
a biplot where all the row points are sqeezed into one another with the column points
more spread out).
1
/
2
1
/
and
C
1
/
2
V
7.6.2 Ordinary PCA biplot of the weighted deviations matrix
Specifying in
cabipl
the argument
ca.variant = PCA
results in an ordinary PCA
biplot of the weighted deviations matrix
R
−
1
/
2
(
X
−
E
)
C
−
1
/
2
. The user can specify, if
required, the argument
scaled.mat = TRUE
for first scaling the columns of
R
−
1
/
2
(
X
−
E
)
C
−
1
/
2
to have unit variances. As an example we again consider the 2007/08 crime
data set. The PCA biplot of
R
−
1
/
2
(
X
−
E
)
C
−
1
/
2
without any further scaling is given in
