Sunday, January 30, 2011

Time Series

I was looking at NPR.org this evening, and came across a fresh example of misunderstood time-series charts. This chart allegedly shows the UN's Food and Agriculture Organization food price index from 2006 to 2010.


To save space on the page (I assume) the artist has broken out each year from 2006 to 2010 as its own separate time series. To read this chart, you must find the 2006 line by looking at the labels at right. Then your eyes must head back to the left, follow the 2006 line from left to right, and then pick up the 2007 line at the left side again. The chart is a left-right-left-right eye workout.

Overall, the graphics are attractive and non-distracting. But the fundamental chart organization is irresponsible. The accompanying article tries to demonstrate that the recent unrest in Tunisia, Egypt, and Yemen can be at least partially attributed to rising food prices. If you look at the 2010 series, this point is valid. But consider the fact that the food index was also above 200 from February to August 2008. This point would be less hidden were the time series given as continuous, sequential years.

It is misleading to show the same information on different time series, because our minds look for temporal and relational patterns. If space is an issue (I can't figure out how, because NPR is a website and a radio network, and this chart will never be physically printed), then the data could be reduced to quarters and compressed. And if only one data series is too boring for the artist or NPR readers, then they should find some related data to display. Starvation rate, staple food costs, or time labels for the financial crisis and Australian droughts and floods.