Correlation and Causation

A correlation is just a pattern: when one number goes up, another tends to go up (or down) too. Ice cream sales and drowning deaths both climb every summer, so they are correlated. But nobody thinks eating ice cream makes you drown. The real driver is a hidden third variable, here, hot weather, that pushes both numbers up at the same time. More people swim when it is hot, and more people buy ice cream when it is hot. The two effects move together only because they share a cause, not because they touch each other.

This matters for data work because it is tempting to read a clean scatterplot or a huge data set and assume the strongest story is the true one. A tight correlation, even across millions of data points, only tells you the two variables move together. It says nothing about why. Sample size can make a correlation look statistically convincing without making it causally true; two million users wearing fast running shoes and running fast marathons could just as easily share a cause like training experience, since serious runners buy serious gear and also run more.

To defend an actual cause-and-effect claim, you need more than a pattern in observed data: a controlled experiment that changes one variable while holding others fixed, a mechanism that explains how one thing could physically produce the other, and ideally the pattern showing up again in independent data sets. This idea connects to how data gets collected and visualized earlier in the unit, reminding you that a well-made chart can be fully accurate and still support a false conclusion if you stop at the correlation.