Correlation vs. Causation
Day 22 of 90 Days DataBytes
If you’ve spent any time around data, statistics, or even scientific research, you’ve probably heard this popular saying:
“Correlation does not imply causation.”
It gets repeated so often that it almost sounds like an unwritten law of Data Science.
But what does it actually mean?
Today, we’ll break down these two concepts and understand why confusing them can lead to poor decisions and incorrect conclusions.
Welcome to Day 22 of 90 Days DataBytes.
Let’s start with the basics.
What is Correlation?
The Cambridge Dictionary defines correlation as “a connection or relationship between two or more facts, numbers, etc.”
In statistics, JMP Statistical Discovery defines correlation as “a statistical measure that expresses the extent to which two variables are linearly related”—in other words, how much they change together.
Simply put, correlation tells us that two variables have some form of relationship. As one changes, the other tends to change as well. This relationship can be:
Positive Correlation: Both variables increase or decrease together.
Negative Correlation: As one variable increases, the other decreases.
No Correlation: Changes in one variable tell us nothing about the other.
Notice something important: correlation only tells us that two variables move together—not why they do.
What is Causation?
The Oxford Learner’s Dictionary defines causation as “the process of one event causing or producing another event.”
In statistics and scientific research, causation means that a change in one variable directly produces a change in another variable.
Unlike correlation, causation cannot simply be observed—it must be demonstrated through carefully designed experiments or rigorous statistical methods that rule out coincidence, bias, and other influencing factors.
In other words, causation answers the question:
“Did this actually cause that?”
Correlation vs. Causation
This is where many people make mistakes (myself included).
Just because two things happen together does not automatically mean one caused the other.
Why?
Because correlation only measures the relationship between the variables being observed. It does not account for other factors that may be influencing both variables.
These hidden influences are known as confounding variables or lurking variables.
Without considering these external factors, it is impossible to confidently conclude that one variable causes the other.
Let’s look at an example.
Suppose we observe in a dataset that beer consumption and weight gain tend to increase together. Statistically, these two variables may show a positive correlation.
Does that automatically mean drinking beer causes weight gain?
Not necessarily.
Other factors could be influencing the relationship.
For example:
A person who drinks more beer may also consume a high-calorie diet.
Reduced physical activity could contribute to weight gain.
Stress or traumatic life events might increase both alcohol consumption and unhealthy eating habits.
Any of these factors could explain the observed relationship.
This is why finding a correlation is only the beginning of an investigation and not the conclusion.
How Do We Establish Causation?
To prove causation, researchers must go beyond simply observing patterns.
They conduct carefully designed experiments, collect evidence, control for confounding variables, and perform statistical analyses to eliminate alternative explanations.
For example, in medical research, scientists do not conclude that a treatment works simply because patients improve after taking it. They perform clinical trials, compare treatment groups with control groups, account for other possible influences, and analyze the results before claiming a causal relationship.
Only after eliminating chance and confounding factors can they confidently establish causation.
Correlation and causation are two of the most important concepts in Data Science and statistics.
Correlation helps us discover interesting relationships worth investigating.
Causation helps us understand whether one factor actually influences another.
Knowing the difference allows us to interpret data correctly, avoid misleading conclusions, and make better data-driven decisions.
Can you think of two things you’ve heard people claim are related, but where the relationship might actually be correlation rather than causation? Share your example in the comments!
Keep learning. Keep building. Keep thriving.
— Michael Ilenikhena
Subscribe to follow the journey as I explore, learn, and build at that intersection—one DataByte at a time.