Creating graphs and charts to explore information
Researchers should explore data before subjecting it to statistical analysis, even when they develop hypotheses ahead of time. Tabular and graphical summaries should constitute a first step in data exploration.
An elegant and concise body of research, recorded in the work of people like William S. Cleveland and Edward Tufte, guides the design of effective graphs.
The research shows, for example, that people perceive differences in line length or position more accurately than differences in angle. However, spreadsheet graphing packages ignore this research, and feature and facilitate presentations like pie charts and three-dimensional options whose visual geometry interferes with their interpretation.
Some resources:
- Naomi Robbins has written an excellent nontechnical text that summarizes what we know about how people decode visual information, and applies it to the design of effective graphs.
- Gary Klass offers specific examples of good and bad and chart design. The companion web site for his 2008 text provides useful guidance on producing charts in Excel while avoiding common pitfalls. He offers commonly heard advice concerning pie charts.
Last updated 3-1-2010 © Dave Thompson