Archive for November, 2007

Quote of the day

I have to share this bit of wisdom from my Good Earth tea bag:

Anything too stupid to be said is sung.


It reminds of the times I’ve tried to write lyrics of my own. I always get stuck because I don’t want to write something too cheesy. But when I read the lyrics of many songs I like without listening to the music, I tend to think, “this is so cheesy”, or “this doesn’t make any sense”.

1 Comment

Visualizing global human development statistics

Thanks to Costa at Sustainable Research for pointing out this amazing data visualization tool. The folks at Gapminder have put together an interactive graph that lets you plot a variety of statistics by country, like “physicians per 1000 people”, and “percent urban population”. After looking at a lot of different relationships (hmm, what happens with % women in the labor force vs. % of government spending on the military?), I’m surprised how few variable pairs have a clear relationship. Most of them look random, or maybe have different trends for different regions. But what begins to become interesting is identifying the outliers, like, who has higher per capita CO2 emissions than the US? Who would’ve guessed Trinidad and Guam? Looking at trends over time is also fascinating. Time runs as a variable-speed animation, so you can step through the decades and watch China’s life expectancy dip during the cultural revolution, and watch Rwanda’s make a startling plummet during the genocide.

Beyond the specific statistics available, this is an amazing tool for visualizing data. Between the x-axis, y-axis, dot color, dot size, and time animation, you can individually select and see 5 dimensions of data at the same time, all with a friendly and effective user interface. Of course, it raises the question of whether people can actually process that much information. I found myself turning one or more dimensions off so as not to confuse myself. Maybe with experience you could train yourself to detect patterns or outliers in a 5-D visualization, but I wonder if you’d be able to see, for instance, a 3-way interaction that wasn’t obvious in any 2-D representation.