Experimental vs. observational studies

Experimental and observational studies are the two fundamental ways of collecting data about a relationship between variables. In an experiment, researchers deliberately impose treatments on subjects (called experimental units) and measure the response. In an observational study, researchers simply record what happens to subjects as they are, without assigning or controlling anything.

The difference is easiest to see with an example. To study whether a new fertilizer increases crop yield, an experiment would randomly assign some plots to receive the fertilizer and others to receive none, then compare yields. An observational study would instead survey farms that already chose to use the fertilizer and compare them with farms that did not — but those farms might also differ in soil, climate, or farming practices.

That contrast explains the central rule: only randomized experiments support cause-and-effect conclusions. Random assignment tends to balance lurking (confounding) variables across treatment groups, so a difference in response can be attributed to the treatment itself. In an observational study, confounding variables remain tangled with the explanatory variable, so researchers can claim association but not causation. Observational studies are still essential when experiments are unethical or impractical — no one can randomly assign people to smoke for decades.

The AP Statistics exam tests this distinction directly in its data collection unit. Expect to classify a described study as experimental or observational, identify confounding variables, and explain why causal language is or is not justified — one of the most reliably tested concepts on the exam.