Graphics Reference
In-Depth Information
sent and how they are collected should be taken into account when choosing scales.
Asymptotic optimality criteria only have a minor role to play.
While Fig.
.
shows the importance of choosing binwidths carefully, it also il-
lustrates some display issues. he horizontal value axis is clearly scaled, but it would
surely be nicer if it extended further to the right. More importantly, the comparison
in Fig.
.
ideally requires that all three plots be aligned exactly and have the same
total area. Not all sotware provides these capabilities.
Graphics should be considered in their context. It may be better to use a scale in
one graphic that is directly comparable with that in another graphic instead of indi-
vidually scaling both. Common scaling is used in one form or another in Figs.
.
,
.
and
.
.
It is one thing to determine what scale to use, but quite another to draw and label
the axes. Too many labels make a cluttered impression; too few can make it hard for
the reader to assess values and differences. (Note that it is not the aim of graphics
to provide exact case values; tables are much better for that.) Tick marks in between
labels oten look fussy and have little practical value. In some really bad situations,
they can obscure data points.
Sorting and Ordering
he effect of a display can be influenced by many factors. When more than one vari-
able is to be plotted, the position or order in which they appear in the graphic makes
a difference. Examples arise with parallel coordinate plots, mosaicplots and matrix
visualizations, all discussed in other chapters. Within a nominal variable with no
natural ordering, the order in which the categories are plotted can have a big effect.
Alphabetic ordering maybeappropriate(astandard default, whichisusefulforcom-
parison purposes), or a geographic or other grouping (e.g. shares by market sector)
mightberelevant. hecategories couldbeorderedbysizeorbya secondary variable.
Figure
.
shows twobarcharts ofthe same data, thenumbers ineach classandin the
crew on the Titanic. he second ordering would be the same in any language, but the
first would vary (for instance, Crew, First, Second, hird in English).
Figure
.
.
Numbers of passengers and crew who travelled on the Titanic, by class, ordered
alphabetically (in German) and by status. Source: Dawson (
)
