I happened across a deal on The Complete Idiot’s Guide to Statistics by Robert A. Donnelly, Jr., Ph.D., at the bookstore. I wanted to brush up a little, so I grabbed it. Unfortunately I’m not even through the first chapter and I’m disappointed. Part of chapter one describes the issues with data presentation. It shows two graphs presenting the fact that the author took some golf lessons and lowered his average score from 98 to 96 from May to July. Here’s his example of a misleading graphic:
He’s correct, that graphic is misleading. Chopping off the bottom of a graph to accentuate the differences should always be questioned. But the alternative he proposes is worse:
First off, at a minimum it takes 18 shots to complete a round of golf, so starting at 0 is just silly. But realistically the goal is par. If par on the courses he plays is 72, the graphic might look like this:
But that still ignores the fact that par isn’t the same from course to course. Maybe in May he was playing on a par 76, and in July on a par 72, in which case he’s actually gotten worse after the lessons. That’s why people comparing their golf skills talk about how many shots over par they average. So the right way to graph this looks like this:
A golfer who is currently averaging 26 over par might only be looking to improve to 15, so it might be reasonable to baseline the graph there, or at 10. You might expect later gains to come harder, so a logarithmic presentation could make sense. At this point I’m just throwing out alternatives — the main point is that baselining the graph at 0 shows a complete lack of understanding of data presentation, or golf, or both.