If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Slate)   Next time some pseudo-intellectual internet blowhard tries to take away your carefully thought-out arguments with that "correlation does not imply causation" yarn, just send them here because YOU WIN   (slate.com) divider line 226
    More: Cool  
•       •       •

27662 clicks; posted to Main » on 02 Oct 2012 at 5:31 PM (2 years ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



226 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | 3 | 4 | 5 | » | Last | Show all
 
ecl
2012-10-02 06:02:05 PM  
Came to crucify this article.

Animatronik: I love TFA. The author says the phrase is overused and flawed, yet reaffirms that it is very much true. Theres not much there, until you realize that people who promote specious arguments using weak correlations have a lot invested in not being required to explain causation very well.

Best example is racism. People point to how some races do better than others financially as proof of inherited differences in intelligence. Clearly this a weak argument but it fits right in with what the author is suggesting. Because Progressives need for ppl to be impressed by arguments based on correlations, so they use similar logic.


Did you just have a seizure?
 
2012-10-02 06:03:02 PM  
Babel Fish, therefore: God.

/puff
 
2012-10-02 06:03:35 PM  

kxs401: RexTalionis: kxs401: Right. Hence "imply," not "prove."

Like I said, maybe "imply" has a different meaning in statistics as opposed to everyday speech. Sort of like how how "consideration" means something different depending on who you're talking with.

You may well be right, but morons on the internet aren't using it that way.



Don't worry too much about the reasoning of a moron on the internet, you don't want to suddenly realize you're letting them write your rules by conforming to a moron's definition of words.
 
2012-10-02 06:03:52 PM  
Usually people say this when someone is concluding a cause from a correlation, usually as part of some sensational news headline -- "Being Republican lowers your IQ" type stuff.

It is usually a valid criticism of the preceding discussion.

Causation certainly will create correlation, but correlation really does cause bad logic that does need active resistance. Heck, most superstition is a result of exactly this error -- things happen together and some weird link of "causation" is suspected. Same with circumstantial evidence. Basically, a lot of actual evil has been caused in the world (burning "witches" etc., blaming Jews, etc.) from exactly this human failing.
 
2012-10-02 06:04:02 PM  
Correlation does not always equal causation, but, without correlation, you can eliminate causation.
 
2012-10-02 06:04:25 PM  

falcon176: pseudo intellectuals="I will repeat verbatim what someone far smarter than me said and I have nothing else to back it up if you try to question it so I'd better write QED on the end which means I win no matter what in latin or some shiat I don't know and if you reply to me I'll just reply with some general insults that I run through a thesaurus 5 times"


major bingo there. people parroting this and that. life is full of them. those that really have something going on don't have time to fark around and piss away time on teh internet.
 
2012-10-02 06:04:39 PM  
I didn't read TFA, but being a logician I thought it might help to mention what "implies" means in (classical mathematical) logic. "A implies B" does not mean merely that the truth of A suggests that B might also be true. It means that it is impossible for A to be true without B also being true. A single example of A and B where A is true but B is false refutes the implication (and of course it is easy to come up with an example where A and B are correlated but A does not cause B.)
 
2012-10-02 06:05:14 PM  
So, after reading that article, I can safely ignore subby's suggestion because? Sweet.

See, the way I've always seen it written is "correlation does not equal causation." Of course correlation may imply causation, but they're not the same thing at all. Reading a two-page article which spends far too many words to state that simple sentence, just so subby can somehow claim that the article refutes that very sentence, well, all I can say is that there are better ways to spend time, I guess.
 
ecl
2012-10-02 06:05:47 PM  

highendmighty: Coolfusis: Oh look, it's another one of the "that thing doesn't exactly mean that!" articles. I'll step aside and let the pedants masturbate furiously over this one.

Hey Doc, don't look now, but your post was pretty pedantic.


i3.photobucket.com
 
2012-10-02 06:05:56 PM  
Slate is making fun of internet cliches.

Next week: Wall Street brokers warn against the dangers of reckless investing.
 
2012-10-02 06:06:08 PM  
Correlation doesn't equal causation.

Causation does equal causation. Don't forget that.

Everything has a cause. Nothing happens just 'cuz
 
2012-10-02 06:06:35 PM  
So, the gist of the article appears to be "the phrase is true, but people say it a lot, so it annoys me".

Slate. Where thoughtless contrarianism is mistaken for journalism.
 
2012-10-02 06:06:52 PM  
The more important point is that the internet is depressing. And blogging might me the most depressing thing about the internet, besides Twitter.
 
2012-10-02 06:06:56 PM  

moran: I didn't read TFA, but being a logician I thought it might help to mention what "implies" means in (classical mathematical) logic. "A implies B" does not mean merely that the truth of A suggests that B might also be true. It means that it is impossible for A to be true without B also being true. A single example of A and B where A is true but B is false refutes the implication (and of course it is easy to come up with an example where A and B are correlated but A does not cause B.)


Thank you.
 
2012-10-02 06:08:40 PM  

blahpers: So, the gist of the article appears to be "the phrase is true, but people say it a lot, so it annoys me".

Slate. Where thoughtless contrarianism is mistaken for journalism.


Slate: where authors are paid by the word.

That's why we received a two-page article confirming, but detailing the author's dislike for, that phrase.
 
2012-10-02 06:08:40 PM  
I think you could safely say that correlation SUGGESTS causation. but "implying" is a more direct form of the link between the two. Saying that it suggests causation merely states that one of the ties between the two may be causing the other. Saying that it implies causation means that one definitely causes the other even if it is not readily apparent.
 
2012-10-02 06:08:54 PM  

uber humper: Everything has a cause. Nothing happens just 'cuz


Not really.
 
2012-10-02 06:09:41 PM  

RexTalionis: I prefer to say "post hoc ergo propter hoc".


That's what I say too. "After this, therefore because of this" The word FALLACY only gets used for LOGICAL fallacies
 
2012-10-02 06:10:05 PM  

likesass: Correlation does not always equal causation, but, without correlation, you can eliminate causation.


Quite right. In other words, causation (of A by B) implies correlation (of A with B).

(At least under some fairly mild extra hypothesis that usually hold in the real world.)
 
2012-10-02 06:15:13 PM  

Donnchadha: kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?

You can probably find a correlation between two seemingly random things, however it does not imply that there necessarily has to be a causal link between them. I could probably correlate levels of beer consumptions with hours of football watched, for example. However, just because people who watch more football might also drink more beer means that watching football causes you to drink beer or that drinking beer causes to you watch football. The societal trend of drinking beer while watching football comes from an external social stimulus, not as an inherent property of beer or football.


You don't know that.
 
2012-10-02 06:16:01 PM  

blahpers: uber humper: Everything has a cause. Nothing happens just 'cuz

Not really.


Everything has a reason?

What are you saying?
 
2012-10-02 06:20:23 PM  

angryjd: Donnchadha: kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?

You can probably find a correlation between two seemingly random things, however it does not imply that there necessarily has to be a causal link between them. I could probably correlate levels of beer consumptions with hours of football watched, for example. However, just because people who watch more football might also drink more beer means that watching football causes you to drink beer or that drinking beer causes to you watch football. The societal trend of drinking beer while watching football comes from an external social stimulus, not as an inherent property of beer or football.

You don't know that.


You don't know that I don't know that
 
2012-10-02 06:20:33 PM  
Regarding the difference between the statements "correlation implies causation" and "correlation equals causation" on which some other posters have remarked, the logical meaning of the latter is "correlation implies causation and causation implies correlation." (At least if one interprets "equals" as "is equivalent to". "Equals" applies to objects and "is equivalent to" applies to propositions.) I think that in any reasonable formuation the reverse implication "causation implies correlation" is true except in trivial cases. Therefore debating the merits of the phrase "correlation implies causation" is equivalent to debating the validity of the phrase "correlation equals causation".
 
2012-10-02 06:21:34 PM  
If the Slate writer sinks, (s)he is not a witch.
 
2012-10-02 06:23:03 PM  
This is another one of those things where people take a math term and think it means the same thing as it does in standard language. See also: "proof" or "equal".
 
2012-10-02 06:23:53 PM  

falcon176: pseudo intellectuals="I will repeat verbatim what someone far smarter than me said and I have nothing else to back it up if you try to question it so I'd better write QED on the end which means I win no matter what in latin or some shiat I don't know and if you reply to me I'll just reply with some general insults that I run through a thesaurus 5 times"


Do you invent the wheel every time you need to drive somewhere?

Try reading.
 
2012-10-02 06:24:03 PM  
Correlation doesn't imply causation. No, no, correlation takes causation out to a nice dinner, buys it a little wine, makes small talk, then quietly drives it home, shakes its hand and never calls again. That's just how it is, and causation will never know what it did wrong or what is wrong with it, but correlation and causation will never be quite the same after that. Sure, they'll pass each other on the street, stand about awkwardly while they catch up on things, all the wile looking for the quickest way out of the conversation. And, still, when they walk away, they wonder for a few seconds about how it all just didn't work out, and what might have been different if it had somehow. Then, one day, when they're both older and much more mature, correlation will finally tell causation the truth, the whole story, but it won't matter because causation already moved on in such alarming and speechless ways that it just doesn't matter anymore, man. It just doesn't matter, and it's probably all for the best. But thanks. Thanks anyway.

And that, son, is the story correlation and causation. Don't ever let it happen to you. If you find our causation, you go after it. You go after it with everything you have, and you make sure it knows how you feel about it and just how far you're willing to go after it. Because, if you don't, someday years from now, you might just find out that you were the cause all along, and not the correlation. And that's the worst thing that can happen in life or in love right there.
 
2012-10-02 06:25:09 PM  
Just because I read an article and then posted "correlation does not prove causation" in the comments section, it doesn't mean the article caused me to type it.

Let this sink in: CORRELATION DOES NOT PROVE CAUSATION!
 
2012-10-02 06:25:10 PM  
the better question is why would any researcher worth their salt ever put out a study that boils down to "we noticed some correlated stuff and we were going to test it out to see if there was anything to it, but then we figured 'eh, feck it, lets go to the bar'. give us more grants"
 
2012-10-02 06:31:19 PM  
Now that was some truly stupid shiat.
I guess that's to be expected when you hire a homeless guy holding a sign by the highway to write for your magazine.

www.slate.com
 
2012-10-02 06:32:58 PM  
"By convention, we call an effect "significant" if the chances of its deriving from a twist of fate-as opposed to some more genuine relationship-are less than 5 percent."

P VALUES DO NOT WORK THAT WAY.
 
2012-10-02 06:33:23 PM  
Yes it provide a hint, but what about when you investigate that hint, and find that there actually is no correlation, just coincidence?

What then?

Ohhhh. Maybe "Coincidence does not equal causation"?
 
2012-10-02 06:34:12 PM  

uber humper: Correlation doesn't equal causation.

Causation does equal causation. Don't forget that.

Everything has a cause. Nothing happens just 'cuz


Maybe, but causation doesn't imply causation, by the same definition of "imply" used in the original. Maybe in mathematics, not in any real life scenario. Which is exactly why that definition of "imply" turns the whole argument worthless.
 
2012-10-02 06:34:52 PM  

kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?


A better way of stating this is that while correlation does not imply causation (see "pirates vs global warming," et al) if you are going to set about looking for causation, correlation is a damn fine place to start.
 
2012-10-02 06:35:20 PM  

trippdogg: Now that was some truly stupid shiat.
I guess that's to be expected when you hire a homeless guy holding a sign by the highway to write for your magazine.

[www.slate.com image 250x153]


But is the guy homeless because he was a philosophy major?
 
2012-10-02 06:37:34 PM  

CreamFilling: uber humper: Correlation doesn't equal causation.

Causation does equal causation. Don't forget that.

Everything has a cause. Nothing happens just 'cuz

Maybe, but causation doesn't imply causation, by the same definition of "imply" used in the original. Maybe in mathematics, not in any real life scenario. Which is exactly why that definition of "imply" turns the whole argument worthless.


I know "imply" means something different outside of mathematics, but could you explain why "A causes B" doesn't imply "A causes B" under your definition of "imply"?
 
2012-10-02 06:37:36 PM  
http://dictionary.reference.com/browse/imply

When people say that "correlation does not imply causation," they are using sense 3.

(It isn't true for sense 1, either, IMHO.)
 
2012-10-02 06:37:51 PM  

KrispyKritter: falcon176: pseudo intellectuals="I will repeat verbatim what someone far smarter than me said and I have nothing else to back it up if you try to question it so I'd better write QED on the end which means I win no matter what in latin or some shiat I don't know and if you reply to me I'll just reply with some general insults that I run through a thesaurus 5 times"

major bingo there. people parroting this and that. life is full of them. those that really have something going on don't have time to fark around and piss away time on teh internet.


I AGREE!!!

Farkin losers...
 
2012-10-02 06:38:50 PM  

AsprinBurn: Correlation doesn't imply causation. No, no, correlation takes causation out to a nice dinner, buys it a little wine, makes small talk, then quietly drives it home, shakes its hand and never calls again. That's just how it is, and causation will never know what it did wrong or what is wrong with it, but correlation and causation will never be quite the same after that. Sure, they'll pass each other on the street, stand about awkwardly while they catch up on things, all the wile looking for the quickest way out of the conversation. And, still, when they walk away, they wonder for a few seconds about how it all just didn't work out, and what might have been different if it had somehow. Then, one day, when they're both older and much more mature, correlation will finally tell causation the truth, the whole story, but it won't matter because causation already moved on in such alarming and speechless ways that it just doesn't matter anymore, man. It just doesn't matter, and it's probably all for the best. But thanks. Thanks anyway.

And that, son, is the story correlation and causation. Don't ever let it happen to you. If you find our causation, you go after it. You go after it with everything you have, and you make sure it knows how you feel about it and just how far you're willing to go after it. Because, if you don't, someday years from now, you might just find out that you were the cause all along, and not the correlation. And that's the worst thing that can happen in life or in love right there.



Keyboard. Now. I'll PM my address
 
2012-10-02 06:40:08 PM  
i.imgur.com
 
2012-10-02 06:40:32 PM  
The waving grass doesn't cause the wind
 
2012-10-02 06:44:20 PM  

xaveth: [i.imgur.com image 400x400]


Reminds me of old joke.

Punchline: Priest says: "Celebrate!? I thought it said celibate!"
 
2012-10-02 06:45:15 PM  

kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?


Another explanation is that C causes A and B. For instance, it's possible that weight gain does not cause diabetes but rather than high levels of blood sugar causes both weight gain and diabetes. There certainly is a correlation between weight gain and diabetes but is the issue running around a 250 pounds when you're 5'3" or is both caused by eating an entire gallon of ice cream every day?

To answer your question, though, of course we should "go looking to find causality". We just shouldn't assume it.
 
2012-10-02 06:46:19 PM  

SineSwiper: Lemme see:

1. Fish oil + heart disease = CORRELATION FAIL
2. Lycopene + cancer = CORRELATION FAIL
3. Echinacea + immune system = CORRELATION FAIL

But, hey, keep farking that chicken about how "correlation implies causation". (Meanwhile, the "supplement" industry would like to rake in millions of dollars from you.)


No, but it does give you a place to start.

Why is it every time I bend over to tie my shoes in the morning I get a knot on my head? A dumb person would indicate "tying shoes causes knots on head". A smart person would use that as a starting point for troubleshooting.

1) Did it always happen?
2) Does it happen to everyone else?
3) Does it only happen in the morning?
4) Does it happen every time I tie my shoes regardless of location?
5) Does it happen depending on the specific shoe I am wearing?

Point is: It implied there is something that needs to be looked at more closely.
 
2012-10-02 06:47:00 PM  

meat0918: Yes it provide a hint, but what about when you investigate that hint, and find that there actually is no correlation, just coincidence?

What then?

Ohhhh. Maybe "Coincidence does not equal causation"?


Some correlations are due only to coincidence. Correlation doesn't mean that things "should" tend to coincide, just that they did tend to coincide when the measurements were taken.
 
2012-10-02 06:47:06 PM  

This About That: Correlation does imply causation. Correlation does not confirm causation.


Increases in texting in the United States has occurred at the same time that traffic fatalities - both in the absolute sense and per mile driven - has decreased. That doesn't mean that there's any sort of causation.
 
2012-10-02 06:48:36 PM  
Anybody want to meet up for some multiple regression later?
 
2012-10-02 06:51:13 PM  

kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?


It's logic along the lines of how in science today you get 'theories' instead of 'laws'. Einstein's theory of relativity is more advanced than the 'law' of gravity, but it's still 'only' a theory. When they find something that's 'more true', it'll be replaced/modified.

Anyways, a Correlation doesn't imply causation(imply being a higher standard of proof), since the question becomes what causes what. Let's take being fat and having heart attacks.
'Some' Possibilities:
1. Being fat causes heart Attacks (A causes B)
2. Heart Attacks cause being fat (B causes A)(one could have a case where most heart attacks aren't noticed)
3. Something else casues both(C causes A AND B, but B takes longer)
4. The study was flawed; not a good random sample
5. It's truly a coincidence.

Basically, you're not allowed to assume that A causes B until you have a theory about HOW A causes B and you've tested it.
 
2012-10-02 06:51:29 PM  

kxs401: I guess I'm just ignorant, because I certainly realize that correlation doesn't PROVE causation, but why doesn't it imply it? If A and B are correlated, possible explanations are that A causes B or B causes A. When we notice that smoking is correlated with lung cancer, why wouldn't we go looking to find causality?


No, you're not, unless my PhD and 18 years as a professional statistical analyst mean nothing. Correlation absolutely implies causation. If there's no correlation between X and Y it says X has no way of predicting Y thus could NEVER cause Y. Any correlation (unless it's obviously spurious) between X and Y could mean changes in X DO cause changes in Y. It's just not certain based on a correlation, alone.

Here's a great example of correlation NOT showing causality: Increases in ice-cream sales are strongly, positively correlated with increases in drowning deaths. However, a THIRD variable - seasonality - is A cause of both. You can bet that seasonality correlates with BOTH ice-cream sales and drowning deaths.

Now...correlation doesn't NECESSARILY MEAN causation. For models that are causal, we use a modeling technique called Structural Equation Modeling - based on covariance (which is just unstandardized correlation) - and this IS a causal model, usually based on some theoretical notion of what is causing what.
 
2012-10-02 06:52:10 PM  
Correlations are very useful information to have. Especially in highly multivariate (real world) systems. Not having causation doesn't change the usefulness or validity of statistical correlations. In fact, not having causation is usually exactly why correlations are useful.

I've noticed the vast majority of people who use the "correlation does not imply causation" soundbite use it in completely inappropriate context. They use it as a complete non-sequiter in a context where no causation was claimed, apparently in the hope someone thinks it means something.
 
Displayed 50 of 226 comments

First | « | 1 | 2 | 3 | 4 | 5 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »
On Twitter





In Other Media


  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.

Report