I heard about micromaps from Naomi B. Robbins during the course that we taught together in New York City in June. She wrote a short article about them a year ago, but I had missed it for some reason. After my day of class, Naomi came to me and said: "You should try to transform some of your examples into 'linked micromaps'". I didn't know what she was talking about. She then showed me a copy of Visualizing Data Patterns with Micromaps, by Daniel B. Carr and Linda Williams Pickle, which I immediately ordered.
I was not disappointed at all. The book is not a general intro to visualization, of course, but an in-depth, concise description of a single graphic form and its varieties. According to the authors, a micromap is "a graphic that links statistical information to an organized set of small maps," and its primary purpose is to "highlight geographical patterns and associations among the variables in your data set."
Micromaps are based on small multiples: When presenting multivariate data for direct comparison, it is usually better to design several tiny graphics rather than "relying on recall of serially presented images" (animation) or on a single, large, ultra-complex, cluttered display.
The most common kind of micromap in the news is the comparative micromap. Comparative micromaps show "sequences of maps indexed by time and other attributes. Emphasis is in comparisons between maps, not within one map." Here you have an excellent one by The New York Times.
In this post, though, I will just focus on linked micromaps, which combine maps with other kinds of graphics. Let me give you some examples. First, a comparison of poverty and education. Linked micromaps, as this example shows, can be nice alternatives to scatter plots and other graphics that represent relationships between variables:
And the one below, which reveals that linked micromaps can use any kind of map, including baseball fields or even (this is mentioned in the book) the human body. They can also contain any kind of graphs, not just dot plots, as you can see here, here, and here.
Visualizing Data Patterns with Micromaps has many quotable passages:
- "Good design seeks to address challenges posed by characteristics of the data, the tasks to be performed, and the skills and limitations of the reader."
- "The graphics designer must find a compromise between simplicity of design that accurately conveys the underlying data patterns while acknowledging human perceptual and cognitive limitations and the complexity of real-world data."
- "Finding a visual representation of data that perfectly resolves the question at hand is not simple matter. The underlying quantitative relationships are rarely simple, and there is a tension between design goals, such as providing context for meaningful interpretation and simple appearance."
An exercise that I expound in The Functional Art and in my classes is a comparison between obesity rates and educational attainment in the U.S. In the book, I mentioned that the graphs that better display the inverse relationship between the two variables, and that let readers accurately rank the states, are the scatter plot and the slopegraph. Facebook's Andy Kriebel transformed my originals into interactive charts, so if you want to see them, read this article. During the NYC workshop, Naomi suggested that a linked micromap based on the same data set could make geographical patterns become visible. I gave it a try last night (forgive me for any mistakes you may find. I designed it really quickly in Illustrator):
Perhaps a micromap is not as effective at displaying the relationship between the quantitative variables as the scatter plot in this case, but it certainly reveals intriguing regional clusters.
Anyway, take a look at Visualizing Data Patterns with Micromaps if you have the chance. It's worth its steep price. Also, visit the authors' websites, as they offer many resources: Daniel B. Carr and Linda Williams Pickle. Finally, The National Cancer Institute has a Java-based application to design micromaps. Check it out.
