*The Elements of Graphing Data*is aimed at those involved with the graphical presentation of data for scientific or technical purposes. Its emphasis is on accurately conveying large amounts of information in such a way as to make decoding easy and effective, rather than on making a dramatic effect on the viewer. This is a more nuts-and-bolts perspective than that of someone like Edmund Tufte and right at the other end from "InfoVis" approaches. The intended audience is scientists and engineers, not people in design, marketing, or advertising.

Cleveland starts with some basic principles of graph construction, looking at ways of making the data stand out, legends and captions, banking to 45 degrees, and scales. Some of this is quite low level, in places almost a checklist of instructions. ("Use a pair of scale lines for each variable. Make the data rectangle slightly smaller than the scale-line rectangle. Tick marks should point outward.") And in some ways people producing graphs these day don't need to worry so much about this, since the plotting routines in statistics software such as R try to "do the right thing" by default.

Software won't, however, tell you what kind of graph to use. The core
of *The Elements of Graphing Data* is a survey of different graphical
methods, describing the most useful kinds of graphs and when they are
appropriate. These include logarithm scales, distributions and qq-plots,
dot plots, different representations of time series, scatterplot matrices,
and coplots, with the use of colour and "brushing" (dynamic graphs)
also touched on. Some of this assumes a basic knowledge of statistics:
Cleveland covers the importance of showing residuals, gives a fairly
detailed explanation of how to fit a loess curve, and looks at ways of
presenting statistical variation.

Finally, part three looks at graphical perception, presenting some of the theory and empirical science behind the earlier advice.

This is all clearly laid out and extensively illustrated, with interesting examples of both good and bad graphs. Some of this may be excessive — do we really need a full page example to illustrate that leaving the exponents off the labels on a log scale is a bad mistake? — but mostly it's helpful.

There's nothing startlingly novel in *The Elements of Graphing Data*,
but it is an elegant presentation of important ideas, which it also
helps to place in their broader context. It doesn't jump out and grab
one the way some of the showier books on graphical presentation do, but
is all the more compelling for its rigour. It is recommended for anyone
whose work involves the graphical presentation of technical information.

Note: thanks to Sean Carmody for giving me a copy of *The Elements of
Graphing Data*.

December 2011

**External links:**-
- buy from Amazon.com or Amazon.co.uk

**Related reviews:**-
- books about mathematics

- books about publishing