Correlation does not imply causation

[moggie]

Gotta love xkcd:

[TZ]

1. They are great. I love xkcd

I gotta say that the phrase "correlation does not imply causation" might be one of the most used phrases of me and others in this group on DU, I suspect..:)

[Heddi]

2. ooh yeah right Mr Big Word Science Talker

Just yesterday I went to the Rite Aid to get some medicine for my cold and on the way out they had these cadburry eggs for buy one get one free, so I got some because it's easter and I love the way it's like a raw egg inside. Anyway. I had had this cold for like 2 weeks and I ate a couple of the cadburry eggs and I fell asleep in a chocolate and insulin driven haze and I didn't get a chance to take my meds but I woke up this morning and my cold was gone, therefore, henceforth, ergo, Cadburry Easter Eggs with the creamy goo inside cures colds. DUH. like everyone knows that.

[LeftishBrit]

3. Thanks - I think I'll put that one on my office door!

[CreekDog]

4. and here i thought sunspots caused stock market swings

:D

?imgmax=800

[immoderate]

5. Correlation does imply causation.

What it doesn't do is PROVE causation. And it could imply falsely.

--imm

[LeftishBrit]

6. Well, maybe.

(sorry, couldn't resist!)

[immoderate]

7. Exactly.

You have to think about what imply means.

--imm

[bluedawg12]

9. Just so I understand

(bear with me ) :)

For example, the statement: There is a positive correlation between smoking and low back pain.

I understand that statement to mean: that there is a higher percentage of people with low back pain who smoke, than among people who don't have low back pain.

It does not say: there is a higher incidence of low back pain in smokers, there may be, but that would be a different analysis and a different statement.

That would be the statement: There is an increased incidence of low back pain in smokers.

In fact, both statements might appear to be intuitively correct. But, only intuitively. Perhaps, smokers are less active, less likely to strain their back, and thus, have less low back pain.

On the other hand, the original statement: There is a positive correlation between smoking and low back pain, may lead to a hypothesis and a search for causation. So, yes, inductive reasoning may follow and lead to research as to the possible cause and effect relationship between smoking and low back pain.

Intuitive reasoning is good, it gives that creative spark and Eureka moment. But, it can also lead to outcome bias.

For example, it may seem intuitively attractive to say: Low back pain increases with age.

However, that is actually not true. The peak incidence of low back pain is around 35 to 50 years of age, after which back pain incidence declines, most likely due to two factors: as people age they become less active and place less demand on their back and second, as the spine ages it loses flexibility ( generally) and there is less excursion of the motion segments.

I think two killers in science are : anecdotals and intuition. They can lead to discovery or to some pretty wrong conclusions.

edit: the letter "I"

BTW- Any comments on my little reasoning exercise is welcome! :) Better I learn it now, than saying it in public someday! LOL.

[immoderate]

10. I'm not sure what you're getting at.

I'm saying that if you noticed that Boy Scouts are more likely to help old ladies across the street, you might infer from the data that it is because they are Boy Scouts that they are helping old ladies to cross the street. The correlation implies causality. But it doesn't prove it.

--imm

[bluedawg12]

11. I agree with you and was thinking about risk factors

and how they may correlate with a condition or disease, which is stilll different than saying it is causative.

[immoderate]

12. Interesting.

I understand what you mean. Like smoking doesn't cause cancer. But it provides favorable conditions for whatever it is that does cause cancer. Here, I think you have to specify what you mean when you say cause. There could be primary and secondary causes.

And then there's implications.

--imm

[bluedawg12]

13. Exactly.

Many people confuse risk factors as being causative. They may or may not be contributory.

And, as you mentioned, there are primary and secondary causes. For example, for low back pain, smoking is linked to back pain, as is aging, as are a whole host of diseases.

I think it's great to connect implications in the sense of researching cause, that "Eureka! I have a theory," moment, in fact there are some studies that smoking may have something to do with blood flow to the disc region, but it's not for sure a direct cause.

Here's another interesting example. Many people believe that their MRI findings are of paramount importance, more important than a basic history and physical, and they assume that an abnormality, such as a few disc bulges are causative of back pain, because discs can be the source of pain for some people. Well, this guy named Scott Boden M.D., did some MRI studies on asymptomatc people in the general population and found that with each successive decade the liklihood of one or more MRI abnormality increased, including disc ruptures, all found in people without a lick of low back pain. So, while an abnormal MRI may correlate with low back pain, it is certainly not causative in every instance.

I'm getting a little tangential, but that statement, "Correlation does not imply causation," really got me thinking about how true that is and how many of us make a leap from correlation to causation, sometimes in error.

[GoneOffShore]

8. I know some people who wouldn't get this cartoon.

Maybe.