Democratizing the tools of visualization? But watch out for the data-as-chartjunk phenomenon
Lee points me to this news article by Anne Eisenberg about the website Many Eyes, which allows people to upload data and visualize it in cool ways. A great idea, but the example used to illustrate the article is just horrible:
It’s a classic example of a graph that looks cool but is just confusing. The data are presented in two dimensions, but the two dimensions don’t mean anything, there are lots of colors with no apparent rhyme or reason, and tons of cool techy-looking details that don’t do anything for me. I’d prefer some bar graphs or (if people can handle it) line plots. It was probably a bad idea that this tool was presented in a newspaper column entitled “Novelties.” I’d rather see graphics be in a column called “Essentials.”
But, hey, everything’s gotta start somewhere. At some point, maybe the fad of cool-looking graphs will go away, and people can settle down to something serious.
P.S. I have no problem with cool—for example, see here for a wonderful applet on the New York Times website—but I think graphmakers have to avoid the data-as-chartjunk phenomenon, in which real data are presented but in a way that makes little sense.
Comments
Do your opinions carry over to the similar-looking graphs used by O’Reilly in their surveys of the computer book market? You can find an example here:
http://radar.oreilly.com/archives/2008/02/state-of-the-computer-book-mar-19.html
Posted by: tom s. | September 1, 2008 07:45 AM
Reminds me of flying over Kansas. Let’s see, that field is corn. Over there is wheat. That must be alfalfa…
Posted by: Jon Peltier | September 1, 2008 08:16 AM
This is not a spatial plot nor are they spatial data; the two axes are correctly ‘meaningless’. The box size shows relative volume, which is apparently very easy to perceive.
Beyond the willy-nilly color coding, the second level of the hierachy is so poorly labeled that its incomprehensible. But the graph style itself isn’t a terrible idea, just the execution.
For ex, here’s a nice 1-lvl treemap showing the origins of customers to a youth hostel http://www.julienbayle.net/complexity/visualization/YouthHostel/
Posted by: Jacob | September 1, 2008 12:28 PM
My pet confusing graph can be seen here:
http://www.axa-equitable.com/MorningstarNewsletter/imgs/story2chart.jpg
I’ve had two financial advisors from two different companies try to hard sell me services with a variant of this graph. They show it to you and ask if you can find any patterns (they get really flustered if you do!). I know it’s supposed to frighten me and make me feel insecure about my own financial decisions, but it really just tells me they don’t know squat about visually representing data.
Posted by: Andi | September 1, 2008 01:17 PM
Jacob,
According to Bill Cleveland (in The Elements of Graphing Data), people are better at perceiving relative lengths than relative areas. Things are particularly bad in the above example because the areas have different shapes and different positions in the meaningless 2-d grid.
Again, I think a barplot or lineplot would be better. The above plot has novelty value but that’s not necessarily helpful in conveying information.
Posted by: Andrew![[TypeKey Profile Page]](http://www.themonkeycage.org/nav-commenters.gif)
|
September 1, 2008 08:18 PM