Randomization

Randomization uses chance to assign treatments or determine observation order. It helps prevent systematic differences and biases from shaping comparisons.

Simple random assignment: A method that assigns each unit independently to a treatment group using chance. This is the basic procedure for assigning treatments without a patterned rule.
Randomized controlled trial: An experiment that assigns participants to interventions by chance and compares their outcomes. Random assignment is the defining feature that supports causal comparisons in this trial design.
Probability: A mathematical measure of the likelihood of events, ranging from zero to one. Chance-based assignment requires a probability model for possible allocations.
Random sampling: A procedure that selects units from a population by chance. Sampling selects who is studied; randomization determines which treatment studied units receive.
Randomization inference: A method of statistical inference based on the assignments allowed by a randomized design. The actual assignment mechanism supplies the reference distribution for testing effects.
Permuted block randomization: A method that randomizes treatment assignments within blocks to keep group sizes similar. It preserves chance assignment while limiting imbalances in group sizes.
Randomized experiment: An experiment in which chance determines which units receive each treatment or condition. This is the general experimental design built around treatment randomization.
Causal inference: Methods for estimating the effects of actions or exposures on outcomes. Randomization is a design strategy for identifying causal effects.
Quasi-experiment: A study estimating intervention effects without random assignment to treatment conditions. It seeks causal answers when chance does not determine treatment assignment.
Design-based inference: Statistical inference grounded in the probability design used to select or assign units. Randomization provides known assignment probabilities for design-based estimates and tests.

Linked from 17 pages