Correlation & Causation

1. Saying “correlation doesn’t imply causation” isn’t enough
2. Humorous introductory/refresher examples of correlations without causation
3. Errors in correlation
   1. Post hoc ergo propter hoc
   2. Cum hoc ergo propter hoc
4. How can we infer causation?
   1. Mills methods
   2. Hills criteria
   3. Some additional considerations
5. Vaccines and autism: A serious example of mistakingly inferring causation
   1. Overview
   2. Vaccine footnotes:
6. Summary
7. Sources

Saying “correlation doesn’t imply causation” isn’t enough

If you follow science and critical thinking pages (and if you follow mine, I’m sorry), then chances are you’ve heard the rule “correlation does not imply causation” repeated over, and over, over, and over, and over again ad nauseum in posts, comments, links, etc.

^{T-Swift for the win. Deal with it. Meme tactically acquired from Pedagogical Memes on Philosophical Themes} 

Let’s face the cliché gorilla in the room: We’ve repeated this mantra so much that we’re no longer “beating a dead horse,” we’re “blending a horsemeat smoothie.” It’s my estimation that (despite the repeated declaration and much to my frustration) people who’ve heard this information still, on occasion, infer causation from mere correlation. What a situation! Granted, given the many flaws of human thinking, such persistent errors are about as surprising as diarrhea after questionable gas station burritos—unpleasant but not unexpected.

Of course, hearing what not to do isn’t the same as knowing what to do. Sure, many correlations are not due to causation, but causation requires some kind of correlation. So the question stands… how do we infer causation? We can’t just go around ignoring every correlation. That could be just as bad.

The short answer is, “It’s complicated.”

The longer answer is, “It’s most definitely cooompliiiiiiiiiicaaaaaated.”

Generally, it is much easier to refute hypotheses that suggest causation than it is to support them. Does this mean the investigative tools of critical thinking (like that of science) don’t work? No, rest assured, the tools work. It just means deeply analyzing hard scientific data to penetrate the unknown and come across the naked truth is not easy. (I’m sure there is a puerile joke in there somewhere). After all, it’s usually easier to destroy a building with explosives than to tediously erect one. But the point is yes: we can zero in causal relationships.

Here, we will do more than just discuss correlations; we will also review how causation can be inferred. Hopefully reading this article will cause more learning. Then again, a marine wrote this, so who the hell knows. First, we will review some humorous but spurious correlations and their logical fallacies, post hoc ergo propter hoc and cum hoc ergo propter hoc. Then, we will explore how we can infer causation. There are many ways to do this, but I will show how two important tools apply in some examples: Mill’s methods and Hill’s criteria. Then, I’ll wrap up with a serious example of erroneously inferring causation. As always, citations are provided because I at least somewhat care about credibility and ackyourassy.

Next

Pages: 1 2 3 4 5 6