📚 Trusted by students, educators & researchers since 2017.
Everyone knows the golden rule of statistics: Correlation does not equal causation.
Just because two things move together doesn’t mean one causes the other. Ice cream sales and shark attacks both spike in July, but eating a cone of vanilla doesn’t make you look tastier to a Great White. They are both caused by a hidden third factor: summer weather.
But if correlation doesn’t prove causation, how do scientists actually prove that anything causes anything?
The answer lies entirely in how you design your study. In statistics, there are only two main ways to gather data: you can be a bystander (Observational), or you can be a puppet master (Experimental).
The Trap of the Bystander: Observational Research
In an observational study, the researcher simply watches the world happen. You hand out surveys, you pull historical data, or you observe behavior, but you never interfere.
The Setup: You survey 1,000 college students. You ask them how many hours they sleep, and what their GPA is.
The Result: You find a strong positive correlation. Students who sleep 8 hours get A’s. Students who sleep 4 hours get C’s.
The Trap: Can you publish a paper saying, “Sleeping 8 hours causes higher grades?” No.
Why? Because you were just a bystander. You didn’t control for Confounding Variables (the statistical term for a “hidden third factor”).
What if the students sleeping 4 hours are also working 30-hour-a-week night shifts to pay for tuition? The job is the confounder. The job causes the lack of sleep, and the job causes the lower grades because they have less time to study. The sleep and the grades are correlated, but one didn’t cause the other.
The Magic of the Coin Flip: Experimental Research
If you want to prove causation, you cannot be a bystander. You must become the puppet master. This is called an Experimental Study, and it relies on one massive, almost magical concept: Random Assignment.
The Setup: You gather 1,000 students. Instead of asking them what they normally do, you flip a coin for every single student.
Heads: You are in Group A. You are forced to sleep 8 hours a night in a monitored lab.
Tails: You are in Group B. You are forced to wake up after 4 hours.
The Result: After a month, Group A has higher grades.
The Victory: You can now confidently say, “Sleep causes higher grades.”
Why does this work? Because of the coin flip.
When you randomly assign people to groups, you accidentally distribute all the “hidden third factors” evenly. That group of students working 30-hour night shifts? The coin flip guarantees that roughly half of them end up in the 8-hour sleep group, and half end up in the 4-hour group.
Random assignment acts like a statistical bleach. It scrubs away the influence of confounding variables. Because the two groups are now mathematically identical in every single way except for the amount of sleep you forced upon them, the sleep is the only possible cause for the difference in grades.
Use this interactive sandbox to see exactly how a Confounding Variable creates a fake correlation, and watch what happens to the math when you click “Randomly Assign”:
Confounding Variable Sandbox
Observational Data
Ice Cream vs. Shark Attacks
Pearson r = 0.00
If Experiments Are Better, Why Do We Still Do Observational Studies?
If random assignment is the only way to prove causation, why doesn’t every scientist just run experiments?
- Ethics: We know smoking causes cancer. But we didn’t prove it with a randomised experiment. It is highly unethical to gather 1,000 healthy people, flip a coin, and force 500 of them to smoke a pack a day for twenty years. We had to rely on massive, highly controlled observational studies.
- Logistics: You cannot randomly assign someone to be “tall,” or “left-handed,” or “grow up in a middle-class neighborhood.” You can only observe those traits.
The Golden Rule: The math doesn’t know where your data came from. An ANOVA will happily spit out a p-value for both an observational study and an experimental study. It is up to you, the researcher, to know what you are allowed to claim.