Subscribe Bookmark


Jun 23, 2011

Tree Maps Come of Age

While I was reading the news the other day, the BBC used a tree map to illustrate Internet traffic in January 2010 for the UK, France, Germany, Italy, Spain, Switzerland, Brazil, US and Australia (from Nielsen). The tree map was a pleasant surprise and a great way to communicate how search sites dominate Internet traffic. Sadly, they didn’t use JMP to make the graph, and while they have some kind of filter there, it’s not as powerful as the Data Filter in JMP. Previously, I had only encountered tree maps as a way to view stock market sectors and stocks in the public sector (on financial Web sites like Smart Money). Perhaps we are witnessing the popularization of a newer kind of graph!

A tree map? What’s that?

Tree maps are at the top of my list for power graphs in JMP. I can’t count the number of times I’ve shown a JMP tree map to customers who had never seen one before, leading me to the conclusion that it’s just not a very “popular” graph. While a lesser known - and underutilized - way to view large amounts of hierarchical categorical data, tree maps are powerful for visualizing information in relation to one or several quantitative variables. Tree maps are space-filling maps with categories represented as different sized rectangles, which live inside another rectangle. Paired with the Data Filter in JMP, the tree map can show - in an instant - everything from stores or product performance to the contents of a hard drive -its original purpose. Inventor Ben Shneiderman created the tree map out of his frustration with sifting through lists of folders, files and file sizes to uncover who was hogging the shared hard drive in 1990 (yes, we had to share hard drives back then!).

I learned that tree maps first appeared in JMP version 6 from JMP Product Manager Jeff Perkinson. A little more digging around on the Internet also turned up an interesting post about the BBC article at Fast Company, which links to the history of tree maps.

Tree maps in JMP

To recreate the BBC graphic, I put the Data Filter and tree map together in a single window with a script (it’s the first script on the JMP data table, which you can find here). I found this much more interesting and controllable than the one on the BBC Web site. For example, the names of the brands are available on all views, and I can select “Media/News” and “Search/Portal” to compare them to each other. I can also filter out the smaller number of total or unique visitors to focus on the major players. I wish the information for country was included in the data set available on the BBC site (it is the last filter, if you want to take a look). But this is a fun way to look at something complex like Internet use in a quick, visual way – leaving no doubt about what most people do when they go for an Internet surf!

A few things to try:

•Hold the control key to get multiple selections with the Data Filter.

•Use the animate function for a fun party trick – found on the red triangle next to the title of the Data Filter!

•Right click next to the “category” in the Data Filter and pick “order by count” to move the singletons out of the way.

•The Data Filter windows are resizable – you can make the lists a little smaller if the list is so long that it shoves the visitor counts off the main page.