Originally Posted by Direckshun
Correlation can be constructed, it's causation that's the bitch.
You're confusing the two.
No, I'm talking about correlation here. You can pick out 10,000 variables and "correlate" them and come up with semi-interesting findings. What is the relationship between soy milk consumption & gun violence? What if the R-squared on that was sky high?
And yes, on the 2nd part, that does lead to the "causation" part. However, I do think that correlations are nearly worthless in many cases because you pick out literally anything - and even these authors in your first post admit there is not any necessary causation here.