Information Technology Reference
In-Depth Information
Ta b l e 9 . 2
Predictions for the three categorical variables: (a)
Rank
,(b)
Gender
and
(c)
Faclty
. Columns of Tables are given only for Prediction regions appearing in biplot
space because no other prediction is possible. The 'diagonal' values giving correct
predictions are shown in bold.
(a)
(b)
Prediction
R2
R3
R5
Total
Prediction
Female
R1
44
0
0
44
Male
Total
R2
180
13
11
204
Observed
Female
205
40
245
R3
95
20
94
209
category
Male
71
412
483
R4
11
6
79
96
R5
11
173
175
(c)
Prediction
F1
F2
F5
F9
Total
F1
59
40
3
48
150
Overall % correct predictions for
Rank
:
F2
56
46
1 0143
(180
+
20
+
173)
×
100
/
728
=
51
.
2%
F3
14
7
3
17
41
F4
24
21
0
14
59
Overall % correct predictions for
Gender
:
F5
14
9
4
2
29
(205
+
412)
×
100
/
728
=
84
.
8%
F6
281 112
F7
53
32
0
35
120
Overall % correct predictions for
Faclty
:
F8
14
32
1
16
63
(59
+
46
+
4
+
57)
×
100
/
728
=
22
.
8%
F9
32
22
0
57
111
Several relationships are evident from Figure 9.9; we highlight the following:
•
females tend to lie towards the lower end of
Remun
,
Age
,
AQual
and
Resrch
;
•
few staff (mainly male) have relatively exceptional large research output, with the
majority of staff members having a value of fewer than one;
•
those staff members with the highest research output come mainly from faculty
F
5, are confined to the rank of full professor and are relatively young.
Figure 9.9 can be reconstructed with the following function call:
X <- Remuneration.data.genbipl.2002
Genbipl(X = X, G = indmat(X[,6]), prediction.type = "normal",
prediction.regions = TRUE, cont.scale = "unitSS", dist.cat =
"EMC", label = F, exp.factor = 1.2, n.int=c(10,10,10,10),
specify.classes = levels(X[,6]), x.grid = 0.005,
y.grid = 0.005)
