Chapter 3 Working with Charts 117
Let’s evaluate what we’ve created. In interpreting bubble plots, the stat-
istician looks for a pattern in the distribution of the bubbles. Are bubbles of
similar size all clustered in one area on the plot? Is there a progression in
the size of the bubbles? For example, do the bubbles increase in size as we
proceed from left to right across the plot? Is there a bubble that is markedly
different from the others? In this plot, we notice immediately that the smaller
bubbles seem to be clustered more toward the left end of the plot. This would
indicate that schools which have a lower percentage of incoming freshmen
that graduate in the top quarter of their class also tend to have a lower ulti-
mate graduation rate. However, it’s also interesting that the bubble represent-
ing Minnesota is slightly larger than its surrounding bubbles indicating that
we probably cannot argue that Minnesota’s lower graduation rate is due to a
lower number of incoming students who graduated in the top quarter of their
class. We would probably have to do further research to discover a reason
from Minnesota’s slightly lower graduation rate.
Breaking a Scatter Plot into Categories
Bubble plots have the problem that it is not always easy to compare the
relative sizes of different bubbles, so another approach we can take is to
divide the universities into categories, plotting universities from different
Figure 3-29
Final
bubble plot
