Pages

Monday, July 4, 2011

Correlation Does Not Imply Causation & Other Fallacies


Among the most common mistakes made by scientific amateurs is the presentation of a correlation as proving a causation. While scientific methodology is quite clear that correlation does not equal causation, does not imply causation, and does not prove causation, many people circumvent this major tenet with frightening regularity. Because of this their statistics are flawed, and their conclusions are usually wrong.

Allow me to explain.

Now, what is a correlation? A correlation is a similarity, a relationship of closeness between two or more variables. Now, what is a causation? The action of causing an effect.

Let's expand upon this more with an example that I was given at the beginning my undergraduate career.
Let's say you're doing fieldwork in the England, measuring stork populations and human birthrates. Now, in the English countryside there is a high population of storks, and a high human birthrate. The correlation is that where you find large stork populations, you find high birthrates. Each variable correlates to the other, positively. Where stork populations shrink, so does the birthrate; where birthrates are high you can be certain to find larger populations of storks.

Now, a person making the mistake of assuming that correlations prove causation could only come to the conclusion that the storks must be bringing people babies! (For you non-Western readers, this is a common stereotype about where babies come from.) Obviously this is an erroneous conclusion, and serves as a classic example of why correlations do not prove causation.

The real reason why stork populations correlate with birthrates is that in the English countryside most people are farmers, and thus need a larger family to help with all the work. And the reason there are large stork populations is because the countryside has less pollution, noise (and other things that storks dislike) than cities or large towns.


Do you understand?

I enjoy the way of reducto ad absurdum when it comes to illustrating important points, but I'll include some legitimate ones as well. Let's keep going with more examples!

1. As ice cream sales increase, the rate of deaths by drowning increases too. Therefore, ice cream causes drowning.

2. Sleeping with shoes on correlates strongly with waking up with a headache. Therefore, sleeping with shoes on gives you headaches.

3. Black people are incarcerated at higher rates than White people. Therefore Black people are more criminal than Whites.

4. Most billboards showing attractive women show White women. Therefore, White women are more attractive than other women.

5. With a decrease in the number of pirates there has been an increase global warming over the same period. Therefore, global warming is caused by a lack of pirates!

Do ya'll understand the concept yet? While the statements seem logically sound, we know for a fact that the conclusions they posit are ludicrous. Which is why correlations can give hints and suggestion as to causes, they don't really tell you much than that there's a relationship between X  Y, while Z, which is the actual cause, is somewhere over yonder being ignored.


Basically, correlations do not imply causation, and frequently lead to absurd conclusions and implications, such as racism, sexism, xenophobia, misandry, misogyny, and outright bigotry, which can be easily contradicted through science, logic, common sense, experiential data, and, of course, venturing outside your personal little bubble.

Unfortunately, many people use correlations to "prove" unprovable things like Black people are dumber than White people.

(Unproveable because, assuming that intelligence can be defined, measured, and quantified, and that all races are pure, you cannot actually physically measure the intelligence of every Black person on the planet, every White person on the planet, and then compare the two in order to conclusively "prove" this claim.)

The above argument is often based on IQ scores, which are said to correlate with intelligence.

Wrong! IQ tests measure how well you take a test. If they measured intelligence, then test-makers would do only as well as their actual intelligence -- instead of frequently acing the tests. Furthermore, correlating how well one takes a test to intelligence does not imply any kind of intelligence, as many people are quite intelligent yet do horribly on tests. Reasons abound: nerves, lack of food, lack of sleep, pressure, negative self-image, etc. There is also no consensus on the definition of intelligence, or how many different kinds of intelligences there are.

This is a basic explanation, regarding the subject. For more information see here


However, assuming that all of the above was not being taken into consideration, that still leaves the impossibility of proving a direct causational link based on the correlation between "Blackness" and intelligence -- in the above example's case, a lower intelligence. Why is this? Because you need to look at the issue holistically. What factors go into measuring Black people's intelligence. How do we account for racism, if any? How do we account for class differences, if any? How do we account for gender differences, if any? What other factors might influence our results besides the aforementioned? What about diet? What about nationality? What about geographic location? What about educational levels of attainment? What about Black/White family life or notions of kinship? What about religion, ecology, cultural adaptations, clothing, music, language, dress, etc and so on.

There are just a few of the questions that legitimate scientists must ask themselves when studying an issue and trying to see if their correlation is merely a suggestive relationship, or an actual implication that X causes Y, instead of Z -- that one causes the other.

In summation, correlations do not imply, prove, or otherwise validate any causational argument. You cannot say Black people are dumber than White people based on test scores, or prove that they are more criminal based on incarceration rates. You cannot say that certain races of women are more attractive because they appear on more signs, billboards, commercials or magazine covers. You cannot determine someone's intelligence based on cranial capacity. You cannot say that men are more logical than women based on the number of men in "hard science" fields. You cannot say that all Jews are rich because they're Jewish based on counting the number of rich Jews. You cannot say any of those things if you are doing science.

But the people who can, and do say these things aren't doing science (much as they would have us believe). So they say and argue these things with impunity. Meanwhile, those of us committed to ethical, methodologically sound, and legitimate science cannot assume that any correlation proves, or even implies -- that is to be a necessary part of -- in the slightest bit any causation. You need to prove this relationship with solid data, a preponderance of evidence, and verifiable, repeatable, methodologically sound arguments.

Once you make a "correlation implies causation argument", you have stopped doing science, and are now just making things up.


I hope this brief lesson will be instructive to those of you dealing with pseudo-intellectuals, scientific racists, academic bigots, and other haters who are wannbe intelligentsia.

Now to those of you reading and wondering who am I to talk about these issues, well... I am an anthropologist.

I study humans.

I study their biology, genetics, culture, history, language, kinship, folklore. I even dig up their leftovers to analyze in laboratories through 4 major sub-disciplines: physical anthropology, cultural anthropology, linguistics, and archaeology. I spend a majority of my time working on things related to this field, and have completed multiple research projects from studying poverty & drug use in the Californian South, to enthomusicologies of the San Francisco open-mic community, and finally the intersection of race, genetics & intelligence, specifically as it applies to American ethno-racial groups.

While admittedly an undergraduate still, I am one semester shy of my baccalaureate, having made Dean's List multiple times throughout my college career in addition to other accolades.

The reason I have briefly described my creds for ya'll here is to illustrate one very important point: I am an expert.

Humble? Probably not, but there you go.

Meanwhile, many of the people discussing these issues, talking about HBD or genetics, culture, racism, etc., are laymen. They are amateurs, and untalented at that. The mistakes they make are elementary (oh yes they are my Dear Watson!) and suffice to prove their lack of expertise or knowledge. (See the above half of this post.)

While one should not appeal to authority, for that is a logical fallacy, one should consider the evidence and expertise of experts over that of people who do not even possess a degree in the field being discussed.


Yet, when confronted with this, many of my detractors attempt to paint anthropology like it is not a "real" science -- see here as well -- comparing it unfavorably to disciplines such as biology, and genetics (despite including fundamental aspects of these fields into itself). Yet they seem just as eager to indulge in the evidence of psychologists, economists, journalists, and even people in ergonomics -- the glorified field of chair-building! Shoot, they even look to forensic anthropologists to confirm their prejudices despite dismissing their peers in the same breathe. Can we say, hypocritical much?

When looking at this trend towards cherry-picking their fallacious appeals to authority, one can only conclude that my detractors discount anthropology merely because the majority of those from the field disagree with their uninformed opinions! See here

(I find this humorous because even those in biology and genetics disagree with their uninformed opinions, no matter how they misrepresent, distort, and otherwise jump to conclusions regarding the data. See here.)

I would like to conclude that while my detractors often begin at a good starting point -- "But no, race does not affect your I.Q. Rather, there is a correlation between AVERAGE I.Q." -- they quickly devolve into race-realism pseudo-scientific racism when they turn that corner into White Nationalism. See here


At this point, one must conclude that the other side is not interested in rational debate, because the other side is not rational.

Which is why posts like these, edifying and elucidating on proper scientific methodology as well as logical fallacies become ever more important when dealing with certain corners of the blogosphere that parlay to the racist, sexist, prejudiced, ignorant, and bigoted minority of American society.


Cheers

3 footnotes:

Mira said...

To be honest, I find this post to be uninspiring, not because I dislike the way you wrote it, but the fact something simple as "correlation does not imply causation" needed to be said (over and over and... over again).

PS- It's generally believed here that stork brings children to their parents, but looking at Dumbo, I guess it's not a local thing.

There's even a children song here, that goes something like: "stork doesn't bring kids, mothers give birth to them". We used to sing it in the kindergarten, along with partisan/anti-fascist songs about WWII war and comrade Tito.

Oh, those were the days...

Zek J Evets said...

Mira,

Haha, I know how you're feeling. But the lack of understanding of basic scientific methodology is actually quite serious among that crowd. So much so that a reiteration of this beginner stuff is a necessity before even bothering to discuss the more important issues.

But now I want to hear more about you waxing nostalgic for comrade Tito. Who is that!? Haha.

Mira said...

Josip Broz Tito was the Yugoslavian president (elected for life). He was the "greatest son of our peoples". He was one of the foundrs of the Non-Aligned nations, and he showed the finger to both East and the West.

In a country without religion, he was the closest to a deity you could get.

http://en.wikipedia.org/wiki/Josip_Broz_Tito

My generation (and a few generation before mine) grew up with the stories of WWII. It was a normal thing to learn in the kindgergarten about the Germans who tortured people and brave boys who were tortured and killed for freedom. Along with the usual children's songs, we had to memorize many patriotic ones, usually about the bombs, war, concentration camps and the partisans.