Hark! the statistical angels sing.

This one is trying to save us from the four horseman of a scientific reproducibility apocalypse. These include “publication bias” that can be set right with meta-analyses including (the often several) unpublished negative trials, “low statistical power” that ushers in wasteful studies, “p-hacking” where investigators run countless analyses and only report the significant ones, and the related “HARKing” where one makes Hypothese After Results Are Known. | Bishop, Nature 2019


Popular Posts