Excellent points have been raised in both the article and the comments (especially the reliance on a single advisor; that's trouble). But one thing that no one

Excellent points have been raised in both the article and the comments (especially the reliance on a single advisor; that's trouble). But one thing that no one is addressing, is the fact that a much of the reproducibility crisis is caused by bad methodologies. Why are PhDs still misapplying statistical tools meant for normal distributions, by using them for power law distributions? That's endemic in the social sciences (see: economics, psychology). Why aren't universities giving all PhD students a proper background in stats & probability?