Information Technology Reference
In-Depth Information
10
10
Resrch (0.85)
Resrch (0.85)
8
8
6
6
4
4
25
25
30
30
2
2
35
35
9
AQual (0.46)
9
AQual (0.46)
40
40
10
10
15
15
45
8
45
8
20
20
25
25
7
50
7
50
0
0
30
30
55
35
Remun (0.64)
55
35
Remun (0.64)
6
6
40
40
60
45
60
5
5
45
50
50
65
Age (0.62)
65
Age (0.62)
4
4
Female
Male
R2 (0.88)
R3 (0.10)
R5 (0.99)
10
10
Resrch (0.85)
Resrch (0.85)
8
8
6
6
4
4
25
25
30
30
2
2
35
35
9
AQual (0.46)
9
AQual (0.46)
40
40
10
10
15
15
45
8
8
45
20
20
25
25
7
0
50
7
0
50
30
30
Remun (0.64)
55
35
Remun (0.64)
55
35
6
6
40
40
60
60
45
5
45
5
50
50
65
Age (0.62)
65
Age (0.62)
4
4
Female (0.84)
Male (0.85)
F1 (0.39)
F2 (0.32)
F5 (0.14)
F9 (0.51)
Figure 9.9
Generalized biplot of Remuneration data. Variables
Resrch
,
Remun
,
Age
and
AQual
are treated as quantitative variables, each centred and scaled to unit sum of
squares. Variables
Gender
(two categories),
Rank
(five academic positions) and
Faclty
(nine different faculties) are treated as qualitative. Inter-sample distances are calculated
using Pythagorean distances for the continuous variables and the EMC for the categor-
ical variables. The quantitative variables are represented by calibrated linear axes. The
calibrations are in terms of the original units using the methods described in Chapter 2.
The length of each axis covers the range of the actually observed values. For the qualita-
tive variables, prediction regions are the counterpart to biplot axes. In the top left panel
we have the samples as points coloured according to gender with the biplot axes of the
quantitative variables. Axis predictivities for the quantitative variables given in brackets.
In the top right biplot we have the prediction regions for variable
Rank
; in the bottom
left panel the prediction regions for
Gender
and in the bottom right panel the prediction
regions for
Faclty
. In the biplots given in the top right panel and in both the bottom
panels the proportion of correct predictions for the respective categories are given in
parentheses.