Information Technology Reference
In-Depth Information
5
30000
300
1.05
32.5
32
6500
10
20000
Group 1; n = 55
Group 2; n = 96
Group 3; n = 39
Group 4; n = 210
Group 5; n = 97
7000
Figure 4.20
Zoomed version of the weighted CVA biplot in top panel of Figure 4.18
revealing clearly the 95% confidence circles about group centroids.
ParEdY
number of years of education completed by the mother for the 1993
data or parent for the 2006 data;
Group
race group - one of Black, Coloured, Indian or White;
Ye a r
1993 or 2006 (initial data set consists only of the 1993 data).
The original data for 1993 formed part of a larger data set describing socio-economic
differences among various race groups in South Africa prior to the democratic elections
held in 1994. The original data set was supplemented by a similar set of data compiled
in 2006. We first consider the following instruments for describing the nature and degree
of overlap among the various groups in 1993 and then also for changes that took place
between 1993 and 2006.
The CVA biplot of the 1993 education data in the top left panel of Figure 4.24
contains the means of the four groups with all the samples added. Since the differences
between weighted and unweighted CVA biplots are negligible, we show only biplots for
the weighted case. The weighted biplot has the following measures of overall quality: in
terms of the canonical variables, 0.9970; and in terms of the original variables, 0.9992.
Axis predictivities are shown in Tables 4.23 and 4.24, with the class predictivities given
in Table 4.25.
