Pet peeve of the day

A picture named 2004_0210-casetuition.jpgI hate graphs that are made intentionally misleading. Graphs should present straight information at a glance whenever possible.

This example is from Fluggart‘s post today about tuition hikes at Case University (a scan from a student newspaper article).

At first glance it looks like tuition (the darkest portion of the bars) has risen about 9-fold over nine years: The tuition bar for 1995-96 is about 8 pixels high and the tuition bar for 2004-05 is about 72 pixels high. But read the numbers and you see that the increase isn’t 900 percent, more like 65 percent.

Back in the fourth grade or thereabout I remember we had a chapter in our math book with a short, very pedagogical case story about graphs. A landlord had announced a rent increase and his tenants found it unreasonable. They presented a graph showing how sharply the rent had risen over the past ten years (picture a narrow line graph, one data point for each year and a steadily rising curve showing the rent). In response, the landlord plotted the rent into a graph showing how the rent had been increasing only very slowly over the same period. His graph included a data point for every month, it was very wide and showed that increases had been very rare, since for almost every month the rent had been unchanged compared to the month before.

We got the math book’s point about scales, of course. But this business with the non-zero starting point for the vertical axis is worse. Isn’t there some rule that says that if your vertical axis doesn’t start at zero, you don’t draw it as connected to the horizontal axis?

Leave a Reply

CommentLuv badge