The Crusade Against Multiple Regression Analysis

I frequently disagree with Nisbett's thought process and conclusions in his other writings, but this analysis of the problem with MRAs is spot on.

MRAs are in my view the last nest of resistant pseudoscience that is still anchored within the mainstream scientific establishment.

As a tool, they're not completely useless if you can't do a controlled experiment... but the misunderstandings of their proper usage case and the endless spurious headlines that result from MRA studies (Tofu being correlated with car accidents and the like) mean that I don't trust most people to utilize them properly.