The following is taken from an article (sorry not yet online) in the most recent Runner's World magazine:

On average the men's marathon performances were 10.71 times slower than their 5k performances, while women's performances were 10.45 slower. In other words, women slowed down less. They seemed to have more endurance. Moreover to quote my [statistician] friend: "While the difference might seem small, it is in fact very statistically significant because the sample size is so large." (Emphasis mine.)

Ugh!!! Very statistically significant? What the #$^@ does that mean? I'm sure the journalist took it to mean something like "very important". But it doesn't mean that at all. In actuality, statistically significant means something more like "under certain assumptions which may or may not be true, the estimated relationship is unlikely to be zero." But the big question is not whether the estimated relationship is nonzero, the big question is whether the estimated relationship large or small. As Deirdre McCloskey would say, "Does it have oomph?" In this case the difference isn't that large; that is, there's no oomph. So who cares if it is statistically significant?

Btw, based on the estimated 10.71 ratio between marathon and 5k PRs, my estimated marathon time is 3:36 (since my 5k PR is 20:09). This turns out to be pretty accurate since my actual marathon time last year was 3:34.

[Welcome also to new co-blogger Ralph Frasca, an econ prof at the University of Dayton.]