Graphics Reference
In-Depth Information
sotware (Murrell,
). It is reasonable to assume that there will be a steady pro-
gression to a situation where even non-experts will be able to draw what they wish.
Whether good graphics are the result will depend on the users' statistical good sense
and on their design ability. Like the quality of a scientific article, the quality of a data
visualization graphic depends on content and presentation. How has the quality of
scientific articles changed since scientists have been able to prepare their own drats
with sophisticated text preparation sotware?
Bad Practice and Good Practice (Principles)
2.6.2
Sometimes it is easier to see what has gone wrong than to explain how to do some-
thing right. Take the simple task of preparing a barchart to display univariate cate-
goricaldata.Whatcouldpossiblygowrong?hebarsmaybetoothin(ortoofat);the
gaps between the bars may be too narrow (or too wide): the labelling of the bars may
be unclear (because it is di
cult to fit long category names in); the order of the bars
may be confusing; the vertical scale may be poorly chosen; there may be superfluous
gridlines; irrelevant
-D effects may have been used; colours or shading may have
been unnecessarily added; or the title may be misleading and the caption confusing.
Doubtless there are even more ways of ruining a barchart.
It is not possible to give rules to cover every eventuality. Guiding principles like
those outlined in this chapter are needed.
And Finally
2.7
he lack of formal theory bedevils good graphics. he only way to make progress is
through training in principles and through experience in practice. Paying attention
to content, context and construction should ensure that sound and reliable graphics
are produced. Adding design flair aterwards can add to the effect, so long as it is
consistent with the aims of the graphic.
Gresham's Law in economics states that 'bad money drives out good.' Fortunately
this does not seem to apply to graphics, for while it is true that there are very many
bad graphics displays prepared and published, there are also many very good ones.
Allseriousdataanalysts andstatisticians shouldstriveforhighstandardsofgraphical
display.
References
Becker, R., Cleveland, W. and Shyu, M.-J. (
). he visual design and control of
trellis display, JCGS
:
-
.
Burn, D. (
). Designing effective statistical graphs, in C. Rao (ed), Handbook of
Statistics,Vol.
,Elsevier,pp.
-
.
Carr, D.B. (
). Designing linked micromap plots for states with many counties,
Statistics In Medicine
:
-
.
