Graphics Reference
In-Depth Information
Choice of Graphical Form
2.4.1
hereare barcharts, piecharts, histograms, dotplots, boxplots, scatterplots, roseplots,
mosaicplots and many other kinds ofdata display.hechoice dependson the type of
datatobedisplayed(e.g.univariate continuous datacannotbedisplayedinapiechart
and bivariate categorical data cannot be displayed in a boxplot) and on what is to be
shown (e.g. piecharts are good for displaying shares for a small number of categories
and boxplots are good for emphasizing outliers). A poorchoice graph type cannot be
rectified by other means, so it is important to get it right at the start. However, there
is not always a unique optimal choice and alternatives can be equally good or good
in different ways, emphasizing different aspects of the same data.
Provided an appropriate form has been chosen, there are many options to con-
sider. Simply adopting the default of whatever computer sotware is being used is
unlikely to be wise.
Graphical Display Options
2.4.2
Scales
Defining the scale for the axis for a categorical variable is a matter of choosing an
informative ordering. his may depend on what the categories represent or on their
relative sizes. For a continuous variable it is more di
cult. he endpoints, divisions
and tick marks have to be chosen. Initially it is surprising when apparently reliable
sotwareproducesareallybadscaleforsomevariable.Itseemsobviouswhatthescale
should have been. It is only when you start trying to design your own algorithm for
automatically determining scales that you discover how di
cult the task is.
In Grammar of Graphics Wilkinson puts forward some plausible properties that
'nice' scales should possess and suggests a possible algorithm. he properties (sim-
plicity, granularity and coverage, with the bonus of being called 'really nice' if zero
is included) are good but the algorithm is easy to outwit. his is not to say that it is
a weak algorithm. What is needed is a method which gives acceptable results for as
high a percentage of the time as possible, and the user must also check the resulting
scale and be prepared to amend it for his or her data. Di
cult cases for scaling al-
gorithms arise when data cross natural boundaries, e.g., data with a range of
to
would beeasy toscale, whereasdata with a range of
to
would bemoreawkward.
here is a temptation to choose scales running from the minimum to the maxi-
mum of the data, but this means that some points are right on the boundaries and
maybeobscuredbythe axes. Unless thelimits aresetbythe meaning of thedata (e.g.
with exam marks from
to
, neither negative marks nor marks more than
are
possible - usually!), it is good practice to extend the scales beyond the observed lim-
its and to use readily understandable rounded values. hereis no obligatory require-
ment to include zero in a scale, but there should always be a reason for not doing so;
otherwise it makes the reader wonder if some deception is being practiced. Zero is
in fact not the only possible baseline or alignment point for a scale, though it is the
most common one. A sensible alignment value for ratios is one, and financial series
