5 ms·

Correlation does not imply causation. https://en.wikipedia.org/wiki/Correlation_does_not_imply_causation#Examples_of_illogically_inferring_causation_from_corre

Correlation does not imply causation. https://en.wikipedia.org/wiki/Correlation_does_not_imply_causation#Examples_of_illogically_inferring_causation_from_correlation https://en.wikipedia.org/wiki/Correlation_does_not_imply_cau...

lightyrs 13y ago
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-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...
eblume 13y ago
No but it is cause for concern.
- kevinthew 13y ago
  This has been confirmed in other studies -- that people with white sounding names do better than those with "black" sounding names on resumes, all things being equal. Our country is incredibly structurally racist is the moral of the story.
TallGuyShort 13y ago
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.
foobarbazqux 13y ago
Causation implies correlation though. That is, correlation is a necessary condition for causation.
- api 13y ago
  As I heard it explained once: correlation does not imply causation, but it does jump up and down and point and gesticulate wildly in its general direction.
  - foobarbazqux 13y ago
    Ha ha. To be fair, every time you hit reply, somebody dies within 3 seconds.
  - a-priori 13y ago
    http://xkcd.com/552/ http://xkcd.com/552/ See the mouseover text.
- doctorwho 13y ago
  https://en.wikipedia.org/wiki/Correlation_does_not_imply_causation https://en.wikipedia.org/wiki/Correlation_does_not_imply_cau... The counter assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as cum hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").
  - foobarbazqux 13y ago
    I think maybe you missed the parent comment?
coldtea 13y ago
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?
- lightyrs 13y ago
  Since we have so many fans of science on this site: http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...
rodgerd 13y ago
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?
- lightyrs 13y ago
  Not at all. However, I also don't then post inflammatory pseudoscientific articles on the Internet with headlines like "White stove caused burns in man who was never burned by black stove." It's the current trend of bad science and lack of journalistic integrity that I'm taking issue with.
  - codygman 13y ago
    Wow... the implication that a white and black stove somehow mirror racism in anyway is laughable. It assumes that the only difference is color. It's disrespectful to the hundreds of years of institutional racism, denigration, and the current state of racial affairs in general.
    - lightyrs 13y ago
      No, that was your inference. The implication was that correlation does not equal causation.
wpietri 13y ago
Correlation is correlated with causation. Which makes it worth paying attention to.
chc 13y ago
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.
ezy 13y ago
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!
- lightyrs 13y ago
  Yes, precisely! It is possible to prove causation, however, this submission does not do so. Please read this primer: http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...
jrs99 13y ago
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.

lightyrs 13y ago

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-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...

eblume 13y ago

No but it is cause for concern.

kevinthew 13y ago
This has been confirmed in other studies -- that people with white sounding names do better than those with "black" sounding names on resumes, all things being equal. Our country is incredibly structurally racist is the moral of the story.

TallGuyShort 13y ago

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.

foobarbazqux 13y ago

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

api 13y ago
As I heard it explained once: correlation does not imply causation, but it does jump up and down and point and gesticulate wildly in its general direction.
- foobarbazqux 13y ago
  Ha ha. To be fair, every time you hit reply, somebody dies within 3 seconds.
- a-priori 13y ago
  http://xkcd.com/552/ http://xkcd.com/552/ See the mouseover text.
doctorwho 13y ago
https://en.wikipedia.org/wiki/Correlation_does_not_imply_causation https://en.wikipedia.org/wiki/Correlation_does_not_imply_cau... The counter assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as cum hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").
- foobarbazqux 13y ago
  I think maybe you missed the parent comment?

coldtea 13y ago

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?

lightyrs 13y ago
Since we have so many fans of science on this site: http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...

rodgerd 13y ago

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?

lightyrs 13y ago
Not at all. However, I also don't then post inflammatory pseudoscientific articles on the Internet with headlines like "White stove caused burns in man who was never burned by black stove." It's the current trend of bad science and lack of journalistic integrity that I'm taking issue with.
- codygman 13y ago
  Wow... the implication that a white and black stove somehow mirror racism in anyway is laughable. It assumes that the only difference is color. It's disrespectful to the hundreds of years of institutional racism, denigration, and the current state of racial affairs in general.
  - lightyrs 13y ago
    No, that was your inference. The implication was that correlation does not equal causation.

wpietri 13y ago

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

chc 13y ago

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.

ezy 13y ago

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!

lightyrs 13y ago
Yes, precisely! It is possible to prove causation, however, this submission does not do so. Please read this primer: http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/?utm_content=buffer69ba9&utm_source=buffer&utm_medium=twitter&utm_campaign=Buffer http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...

jrs99 13y ago

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.