Information Technology Reference
In-Depth Information
Graphs should be simple, with no more than a few plotted lines and a minimum
of clutter. If the graph is being used to demonstrate variation in output values for a
range of input values, the horizontal or
x
-axis should be used for the parameter being
varied, or the input; the vertical or
y
-axis is for the function of the parameter, or the
output. Plotted lines of discrete data should always have points marked by distinctive
marks such as circles, boxes, or triangles.
Consider using greys, colours, or line thickness rather than dots and dashes to
distinguish between lines. If you use shades of grey to distinguish different elements
in the graph, ensure that the shades are sufficiently distinct; lines in lighter grey
sometimes need to be a little thicker than other lines. Greys are preferable to cross-
hatching for filled-in areas in a figure, as the latter can create the optical illusion of
shimmering and does not always print or photocopy well.
Colours are more eye-catching than are greys, but, in a graph, do not necessarily
communicate better. First, some journals do not print in colour. Second, colours can
render inconsistently in different media; for example, a vivid yellow on a screen
may be almost invisible on paper. A related issue is that some colours are more
conspicuous than others, in particular red—a reader may not even notice information
in mild colours if a bright red is present. Third, greys are emotionally neutral, and
thus don't carry the subconscious messages that colours can. In a sales pitch, colour
is used to make implicit statements about products or methods, and to persuade
through visual excitement. In a write-up, where the goal is persuade through logic
and evidence, such an approach is inappropriate.
Minimize use of unnecessary elements and remove all decoration. Are the sec-
ondary ticks on the axes useful? If not, discard them. Is a legend necessary? If not,
remove it, and label the lines directly. Do the captions have to be in a large font? If
not, diminish them. Are the fonts and font sizes different to those of the rest of the
paper? If so, change them. Axes should be inconspicuous; ink should be used for
data, not dressing. Gridlines and boxing are other forms of unnecessary ornamen-
tation. Secondary marks, such as axis ticks, should be a little lighter than the other
elements. The lines and other elements should be of similar weight—don't mix a
large, bold font with lightly drawn lines, for example.
Many of the commonly used graphing tools provide features that are only rarely
of value; worse, some of these features are invoked by default.
1
Poor versions of a
graph are shown in Fig.
11.1
, with revisions of it in Fig.
11.2
. A slightly more complex
Note, though, that these examples illustrate just a few aspects of design of graphs.
With the breadth of kinds of data and result reported by researchers, and also the
breadth of tools available for interpreting and presenting data, it would not be sensible
to attempt to present a comprehensive set of examples; instead, these graphs are
1
Some widely used tools have truly strange default settings, currently including unusual colours
for labels and legends, massive fonts on the axis labels, and dense grids across the background.
When a paper includes a graph produced in that way, the immediate message is that the author is
uninterested in trying to communicate well.
