Having just returned from the Gordon Research Conference on Chemical Education Research and Practice, I can attest to the central role that statistics plays in chemical education. "Of course", every published experiment in chemical education results in a statistical improvement, so it is amazing that all of those three, five, or ten percent improvements have not, so far, changed the overall performance of our students. Statistical methods are essential to prove that one teaching method is superior to another, and the differences are not generally very large. This situation can lead one astray, as the authors show occurred in the published "proofs" by S. Kanazawa that gender ratios are significantly affected by the attractiveness, physical stature, occupations, and tendency toward violence of the parents. A cautionary tale for educational researchers.
Andrew Gelman and David Weakliem