Multiple correspondence analysis - Understanding Biplots - page 401

the Burt matrix. The category centroid of the females in Figure 8.16 is well separated
from that of the males. Furthermore, the category centroid of the females is closer to the
category centroids of staff in the lower income deciles, the younger staff, the more junior
staff members, and those staff members with the smallest research output. This stands in
sharp contrast to the category centroid of the males. When the first dimension (horizontal
scaffolding axis) in Figure 8.16 is replaced by the third dimension, the positions of
the category centroids of the males and females almost coincide. We note also large
differences in the positions of the category centroids of the various faculties, with F 6
quite distant from the rest (in both Figures 8.16 and 8.17). The relative positions of the
CLPs in the Burt matrix based biplot of Figure 8.18 are similar to those in Figure 8.16
but they occupy more peripheral positions relative to the individual sample points.
In Figures 8.19 - 8.21 we give categorical PCA biplots of the Remunera-
tion.cat.data.2002 data with all variables taken as ordinal categorical except
Faclty and Gender . In these biplots several devices are illustrated for enhancing

Resrch

Remun

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10

Res0 Res2 Res4 Res6

Age

Rank

A1 A2 A3 A4 A5 A6 A7

Rnk1 Rnk2 Rnk3 Rnk4 Rnk5

Gender AQual

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9

Female Male

Faclty

F1 F2 F3 F4 F5 F6 F7 F8 F9

Figure 8.22 Final optimal z-scores for each variable, suggesting that some categories
should be lumped together.