Discussion about this post

User's avatar
Emiel de Jonge's avatar

I think considering the moral panic around this, that a minimum of r = 0.2 is to lenient. Considering how often such correlations are spurious or practically meaningless. This sets up the stage for cherry pickers (which are all people with a vested interest, including many researchers) will just misuse these to make statistically unsound claims about the strength of the finding, or if 20 from 100 studies finds a similar finding with r = 0.15 - 0.22 than they can claim "consensus" in psychology, considering how standards of quality research has been lowered in psychology. It does not matter if 80 of the 100 say something different.

I'd say the minimum should be 0.3. If you simulate the statistics and see what the plots show in linear relationships I'd presume very few people with statistical knowledge would claim an r = 0.2 as something useful.

Expand full comment

No posts