No they don’t. This is what people who don’t deeply understand statistics but want to sound smart say all the time. Correlation implies that you can consider causation, but doesn’t prove it. But people use that phrase as if correlation disproves causation. That’s plain stupidhttps://twitter.com/LeviABx/status/1244970629926998016 …
-
-
Yes, exactly. I’ve seen a few exceptions but that phrase has almost always meant that the person doesn’t know what they’re talking about.https://twitter.com/ericrweinstein/status/777401909376225282 …
Show this thread -
By the way, the original tweeter also says he was joking! I'm keeping this thread up because that phrase is indeed so common.https://twitter.com/LeviABx/status/1246508326172704769 …
Show this thread -
For weeks, I've been hearing "but correlation does not imply causation" re:masks despite that in this case, yes, yes, it does: we have coverage and a control (Japan, screwing up everything but masks and it's not Lombardy or NYC); analytic reasons (virus); priors (good studies).
Show this thread
End of conversation
New conversation -
-
-
Somebody’s been catching up on their Khan Academy videos while staying at home
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Correlation and causation don't always go together, but if two things don't correlate, there's no causal connection, so correlation obviously informs us about causation to an extent.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Smoke doesn't *prove* fire - but it does *strongly suggest it*.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Or they're just pointing out that the person making the claim is trying to imply causation when they only have evidence of correlation, which happens A LOT.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
My brain just spent a few seconds doing the truth tables for A implies B, B implies A, not A implies not B, not B implies A.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
But causal inference methods clearly show that with selection and/or adjustment biases even cause and effect can show no correlation, such as when you adjust for an intermediate variable that blocks the causal effect.
-
And if you adjust for a collider you can find correlation between to variables that do not have a causal connection. So you really need a causal DAG or a careful application of potential outcomes to say much here.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.