Information Technology Reference
In-Depth Information
0.1
0.15
04
0.1
07
RAC
03
05
Gaut
AtMr
08
02
0.05
06
CrJk
CmRb
0.05
0.02
KZN
InAs
DrgR
DrgR
AtMr
CmRb
0.02
InAs
PubV
CmAs
0.2
0
0
WCpe
0.01
BRs
0.1
CmAs
Arsn
Mrd
Rape
BNRs
NW s t
Mpml
FrSt
Limp
0.005
NCpe
0.01
0.05
ECpe
0.005
0.02
AGBH
0.01
Arsn
0.02
0.1
Figure 7.23 Pearson residuals case A CA biplot of the 2007/08 crime data with the
interpolated points for the Gauteng province for the periods 2001/02, 2002/03, ... ,
2007/08. For this biplot it was necessary to set in cabipl the argument exp.factor =
1.5 to allow the appearance of all interpolated data points in the display.
two-dimensional approximation to the row chi-squared distances. From Figure 7.31 we
see that products A and C are close together, as is to be expected from the similarity of
their row profiles. But this does not reflect the commercially important information that A
has much lower ratings than does C . The doubling process is designed to rectify this. The
result is shown in Figure 7.32, where A and C are now the most distant pair of products,
reflecting the actual ratings. In these figures, we have included axes representing the
attributes; these have no particular interest, but we note a pattern of clustering which we
discuss further when considering the column chi-squared distances.
Figures 7.31 and 7.32 are constructed with the following calls to cabipl and
cabipl.doubling , respectively:
Search WWH ::




Custom Search