A paper that I read years ago (via a Cosma Shalizi post on the file drawer problem) and then totally failed to find when I wanted to re-read it. Up there with the "No Free Lunch Theorem" as being simultaneously blindingly obvious and something that I'd never thought about.
It discusses exactly what the title says. A highly significant result may be indistinguishable from an insignifcant one, statistically speaking.
Comparisons of the sort, “X is statistically significant but Y is not,” can be misleading.
No comments:
Post a Comment