Experimental design

Experimental design is the planning of studies to test hypotheses while controlling bias and uncertainty. It specifies how treatments, participants, measurements, and comparisons are arranged.

Randomization: A procedure that assigns units to conditions by chance rather than choice. Chance assignment balances known and unknown influences across experimental groups.
Causal inference: The study of methods for estimating cause-and-effect relationships from data. Experimental plans aim to support causal claims, not merely describe associations.
Randomized controlled trial: A study that randomly assigns participants to an intervention or a comparison condition. It is a common design for testing whether an intervention causes an outcome change.
Observational study: A study that measures exposures and outcomes without assigning participants to treatments. Unlike an experiment, it relies on naturally occurring differences rather than planned allocation.
Control group: A comparison group that does not receive the treatment or receives a standard alternative. Its outcomes help estimate what would have happened without the experimental treatment.
Treatment effect: The difference in outcomes attributable to a treatment under a specified comparison. A design must define which treatment contrast its data can estimate.
Crossover study: A study in which participants receive multiple treatments in a specified sequence. Within-person comparisons can reduce variability when treatment effects do not persist between periods.
Before-and-after study: A study that compares outcomes measured before and after an intervention in the same setting or group. Without a concurrent comparison, time trends can be mistaken for treatment effects.
Replication: The repetition of an experiment or measurement to assess consistency and variability. Repeated observations distinguish stable treatment effects from random fluctuations.
Confounding: A distortion of an exposure–outcome association by a common cause of both. Uncontrolled differences between groups can make an apparent treatment effect misleading.

Linked from 28 pages