Graphics Reference
In-Depth Information
Figure
.
.
Procrustes rotation of metric onto non-metric scaling for rail cost
Minimizing S can be di
cult because of the possible large number of local minima.
Various algorithms have been proposed, for example Guttman (
), Hubert and
Arabie (
,
) and Lau et al. (
). Here we use the algorithm by Guttman
(
) for the following example.
he data used to illustrate this method are ratings for World War II politicians
and are the lower triangle given by Everitt and Dunn (
). Figure
.
shows the
resulting unidimensional scaling obtained.
For clarity, the points have been plotted on a diagonal, which represents a linear
axisstarting fromMaoTseTungat
.
andendingat Mussoliniat
.
.hecountries
the leaders represent have also been added. What is interesting is that Stalin is most
closely identified with Hitler and Mussolini as opposed to his UK/US World War II
allies.