Information Technology Reference
In-Depth Information
Ta b l e 4 . 1 5
Remuneration data: within-class sample predictivities for the first four and
last four samples (weighted and unweighted CVA).
Weighted CVA
1
2
3
4
725
726
727
728
Dim_1
0.1530
0.5511
0.4751
0.4009
0.0494
0.1624
0.0028
0.3043
Dim_2
0.2323
0.5518
0.5650
0.8166
0.0670
0.4510
0.0042
0.3060
Dim_3
0.7094
0.7407
0.7884
0.8338
0.0980
0.6190
0.0055
0.3060
Dim_4
0.9646
0.9938
0.9997
0.9809
0.5536
0.7330
0.7394
0.4108
Dim_5
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
Unweighted I CVA
1
2
3
4
725
726
727
728
Dim_1
0.1530
0.5511
0.4751
0.4009
0.0494
0.1624
0.0028
0.3043
Dim_2
0.3126
0.6097
0.4794
0.5146
0.0757
0.2981
0.2204
0.3219
Dim_3
0.4500
0.8705
0.8620
0.9909
0.2665
0.5249
0.9377
0.3219
Dim_4
0.9709
0.9760
0.9970
0.9926
0.6204
0.8792
0.9904
0.5939
Dim_5
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
Unweighted I -
K
−
1
11
CVA
1
2
3
4
725
726
727
728
Dim_1
0.1530
0.5511
0.4751
0.4009
0.0494
0.1624
0.0028
0.3043
Dim_2
0.1895
0.5534
0.5484
0.6952
0.2697
0.6369
0.0096
0.4169
Dim_3
0.2551
0.5580
0.5814
0.6953
0.2934
0.7123
0.2357
0.4315
Dim_4
0.4490
0.8590
0.9009
0.9941
0.3160
0.7649
0.9898
0.4943
Dim_5
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
Finally, we compare the 2002 data with the corresponding data in 2005 in the form
of the CVA biplots given in Figure 4.16. A comparison of these biplots shows that the
gender gap regarding all five variables persisted in 2005. The remuneration gap even
shows an increase in absolute terms. It is also of interest that both groups show an
increase in research output in 2005 as compared to 2002.
4.10 A five-class CVA biplot example
We introduced the copper froth data in Section 3.8.4, where we constructed PCA biplots.
Here we exploit the group structure imposed on the data using the various forms of CVA
introduced in Section 4.2. In a call similar to that given below, to
CVA.predictivities
,
the output listed in Tables 4.16 - 4.22 is obtained. Figure 4.17 contains several snap-
shots of a three-dimensional CVA biplot of the Copper Froth data obtained with
function call
CVAbipl(CopperFroth.data[,1:8], G =
indmat(CopperFroth.data[,9]), weightedCVA = "unweightedCent",
dim.biplot = 3, factor.x = 1.5, factor.y = 1.2,
factor.3d.axes = 0, specify.classes = 1:5)
