Common Errors in Statistics (and How to Avoid Them), 4th Edition

KISS

Keep your graphs simple but complete. A particularly horrific example is located at http://www.aptech.com/3dcontour2.html with a copy at http://statcourse.com/research/sillygraph.jpeg.

Its flaws include all of the following:

1. The unnecessary shading and a false third dimension provide a distracting optical illusion as the cube appears to flick toward and then away from the viewer.

2. The unnecessary third dimension is meaningless as a single continuous variable (burn time) is plotted against a single categorical variable (fabric type).

3. The unnecessary color coding in the bars is distracting; it duplicates the information one can read directly from the Y axis.

4. Do the disks near the top of each bar point to the true burn time? Or does the burn time correspond to the top of the thin bar or the fat bar?

5. I am guessing that the categories on the left correspond to synthetic fabrics and those on the right to natural fabrics; still, a further label would have been helpful.

6. As the graph is separated from its descriptive context, a label providing the details of how burn time was determined is called for.

7. The one bit of seemingly relevant labelling, “average of three samples,” is accompanied by a distracting orange blob.

Rules for avoiding similar catastrophes in your own work are provided in the sections that follow.