And yet it's how we imply causation all the time.

In real life, were we can also draw from other experiences and make our "experiments" more targeted (as opposed to some scientific study involving obscure correlations), correlation is the best measure of causation.

So, while it might be the correct pedantic thing to say, in this correlation does very much imply causation. It's not like there are a lot of others, mysterious, factors at play here.

Besides blanket epistemological statements, how about we also do a little thinking on individual cases and how it applies to them?

The problem with this mantra is that it is technically true, but, as a negative, leaves out the rest of the world. For example, Correlation also does not imply non-correlation. This is vacuous, of course, but it illustrates the problem with using this kind of shorthand to quickly dismiss something rather than investigate it.

In truth, correlation does suggest causation, but not necessarily in the way one might expect. The WP article you linked goes into this.

All too often, this canard is trotted out to avoid thinking about where the causation might actually lay, rather than taking something seriously. It is precisely this that the shorthand is meant to prevent! Correlation does not imply causation, therefore you have to think carefully about the causation chain when considering two correlated events -- you can't just assume the causation isn't elsewhere, and you most definitely cannot assume it does not exist at all!

Causation implies correlation though. That is, correlation is a necessary condition for causation.

Agreed, but it does sound like she was pretty careful in which factors in the experiment were allowed to vary, and unless she's exaggerating the numbers (which is hard to do when you enumerate specifics), that's a pretty significant change. I would be more sceptical if someone pointed out factors that might have varied anyway.

edit: The comment about 'profile freshness' is intriguing, and possible an example of this.

Thanks for the link. I haven't heard this one yet in Internet forums. Now I can win arguments by throwing it around with abandon.

No but it is cause for concern.

Look, I understand many of you are upset by this, however, there are ways to prove causation — it's just a lot more complicated than this.

Here's a primer:

http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...

If we're using strictly logic jargon, the statement "correlation does not imply causation" is true.

If we're using the normal English definition of "imply," then the statement is very often false.

Correlation is correlated with causation. Which makes it worth paying attention to.

So if you burn your hand on a stove element, do you keep your hand there until you've derived a theory of how the element might be burning it from first principles, all the while chanting your mantra?