Visualizing the World

The increased ubiquity of the internet has given more and more people greater access to massive amounts of diverse data.  As the amount of data has steadily increased, so too has the processing power of the machines capable of processing this data.  However, Producing reams and reams of data is not enough – you need people who are capable of analyzing this information and visualizing in ways so that other people can also make sense of it.  Last year, the New York Times foretold that statistics will be the next great emerging field, quoting Google’s chief economist Hal Varian as saying, “The sexy job in the next 10 years will be statisticians.  And I’m not kidding.”

Various companies have developed a number of powerful data visualization tools, and the New York Times in particular has been singled out for praise from the blogosphere for its innovative web graphics.  Nevertheless, before descending into the rabbit hole and trying to visualize anything and everything, it’s important to first understand the design principles that make certain data visualizations more or less effective.  The September 1984 edition of the Journal of the American Statistical Association published a paper by William S. Cleveland and Robert McGill that reported on experiments to determine which visualization methods are most easily understood by an audience.  The authors studied what individuals can decode most accurately and ranked the approaches in the following list:

  1. Position along a common scale e.g. scatter plot
  2. Position on identical but nonaligned scales e.g. multiple scatter plots
  3. Length e.g. bar chart
  4. Angle & Slope (tie) e.g. pie chart
  5. Area e.g. bubbles
  6. Volume, density, and color saturation (tie) e.g. heatmap
  7. Color hue e.g. newsmap

It would be very valuable to produce an updated version of this study.  Have things changed in the last 25 years?  Do these trends hold true for all segments of the population?  Perhaps some individuals with special training, such as soldiers, are more attuned to certain visualization methods.   If anyone knows of any additional studies along these lines, please send them our way.  In addition, Strategic Social is going to be attending a data visualization course led by Edward Tufte in Crystal City today; if you’ll also be there, come say ‘hi.’