3 ms·

Correlation =/= causation. Can't be said often enough.

zeofig 8y ago
Came here to post this.
onion2k 8y ago
Sometimes correlation isn't causation, but if we just say "correlation =/= causation" every time something happens we'll never discover anything. If two things are correlated it's at least worth investigating further, and quite often turns out to be a causal link.
- adrianN 8y ago
  I'm reasonably sure that the set of things that are correlated is much, much larger than the set of things that have a causal relationship. Check for example http://www.tylervigen.com/spurious-correlations http://www.tylervigen.com/spurious-correlations
  - onion2k 8y ago
    The number of things that aren't correlated is waaaaaaay bigger. It's still worth investigating when there's a possible link.
- AznHisoka 8y ago
  And its definitely worth examining if there is a rational, simple hypothesis that implies a relationship between the two. Here, that is clearly the case: could what you think about something impact your health? It doesnt seem like such a stretch to believe so (see placebos, etc) On the other hand, if you found a correlation between butter exports in India and mortality rates, you probably shouldnt dig further into whether there’s a causation relationship. It fails the Occam’s Razor test missrably.
- slx26 8y ago
  You are right, but in my opinion the page/writer completely fails when they don't explain and clearly state that the possibility of "correlation =/= causation" exists. And it fails because we lose 80% of the people to blind belief towards what the article says, and from the other 20%, 15% are saying "correlation =/= causation", and a 4% like you are saying in response: "sure, but that's something we need to investigate to tell". I understand that the balance between science and selling the science is hard, but it can be done, and imho they could be doing it better.
GoToRO 8y ago
But... correlation includes causation so if you want to find the causation you should do a quick test to see if you have correlation. Because is cheaper. THIS can't be said often enough. What you said I heard a ton of.

Came here to post this.

Sometimes correlation isn't causation, but if we just say "correlation =/= causation" every time something happens we'll never discover anything. If two things are correlated it's at least worth investigating further, and quite often turns out to be a causal link.

adrianN 8y ago
I'm reasonably sure that the set of things that are correlated is much, much larger than the set of things that have a causal relationship. Check for example http://www.tylervigen.com/spurious-correlations http://www.tylervigen.com/spurious-correlations
- onion2k 8y ago
  The number of things that aren't correlated is waaaaaaay bigger. It's still worth investigating when there's a possible link.
AznHisoka 8y ago
And its definitely worth examining if there is a rational, simple hypothesis that implies a relationship between the two. Here, that is clearly the case: could what you think about something impact your health? It doesnt seem like such a stretch to believe so (see placebos, etc) On the other hand, if you found a correlation between butter exports in India and mortality rates, you probably shouldnt dig further into whether there’s a causation relationship. It fails the Occam’s Razor test missrably.
slx26 8y ago
You are right, but in my opinion the page/writer completely fails when they don't explain and clearly state that the possibility of "correlation =/= causation" exists. And it fails because we lose 80% of the people to blind belief towards what the article says, and from the other 20%, 15% are saying "correlation =/= causation", and a 4% like you are saying in response: "sure, but that's something we need to investigate to tell". I understand that the balance between science and selling the science is hard, but it can be done, and imho they could be doing it better.

GoToRO 8y ago

But... correlation includes causation so if you want to find the causation you should do a quick test to see if you have correlation. Because is cheaper. THIS can't be said often enough. What you said I heard a ton of.