Information Technology Reference
In-Depth Information
The complete function call used to construct Figure 8.23 is:
CATPCAbipl (Xcat = Remuneration.cat.data.2002[,c(2,3,4,5,
6,8,9)], ax = 1:7, calibration.size = 1.2, calibration.pch =
15, calibration.label.size = 0.7, calibration.label.offset =
rep(0.3,7), calibration.label.pos = c(2,1,1,2,1,1,4),
c.hull.n = 10, class.cols = c("orange","cyan"),
class.pch = rep(2,nrow(Remuneration.cat.data.2002)),
class.vec = Remuneration.cat.data.2002[,6], exp.factor = 1.75,
drawbagplots = TRUE, factor.type = c(rep("ord",4),
"nom","ord","nom"), line.type.bags = rep(2,2),
nom.col = UBcolours[1:9], offset = c(0, 0.3, 0.15, 0),
ord.col = rep(4,12), orthog.transy = c(-0.35, -0.60,
0.16, -0.31, -0.3, 0.15, -1.4), plot.samples =
1:nrow(Remuneration.cat.data.2002), pos = "Hor",
reverse = TRUE, samples.col = rep("blue",
nrow(Remuneration.cat.data.2002)), samples.size = 0.75,
select.origin = FALSE, specify.bags = levels(Remuneration.
cat.data.2002[,6]), w.factor = 1.25)
The biplot in Figure 8.23 provides us with a display that mimics an ordinary scat-
terplot: all data points are shown together with information on all variables using biplot
axes that do not distract attention from the sample points. Calibrations on axes are given
in terms of CLPs. Axes for nominal variables are 'calibrated' in different colours, indi-
cating no natural ordering of the categories; axes for ordinal variables are 'calibrated'
in the same colour but with increasing width indicating ordered category levels. Blasius
et al
. (2009) give a further example of this type of categorical PCA biplot.
Although the 0.95-bags of the males and females in Figure 8.23 show a large degree of
overlap, it is also clear that the males tend to occupy a position indicating higher-ranked
positions than the females, higher qualifications, more advanced age, higher remuneration
and more research output. Overall it can be concluded that the biplot shows evidence of
a gender gap. The changes in the dynamics of this gap can be studied by constructing a
similar biplot of the 2005 data. We leave this as an exercise to the reader.
In the next chapter we discuss how to construct biplots with some continuous and
some categorical variables.
