On this page
Identify the population question
Random selection from a defined frame can support a design-based estimate about that frame’s population. It does not by itself assign a treatment or establish which difference was caused by an intervention.
Identify the intervention question
Random assignment allocates a treatment among participating units under a specified experiment. A study can randomly assign volunteers even though the volunteers were not randomly selected from the wider population. External population coverage and treatment assignment are separate design questions.
Read the two design dimensions
| Study | Selection | Assignment |
|---|---|---|
| Random firm survey with no intervention | Random frame selection | None |
| Volunteers randomly allocated a training exercise | Volunteer entry | Random assignment |
| Randomly selected staff then randomized training | Random selection | Random assignment |
Report the design that supports the conclusion
An observed association in sampled data does not become causal solely because selection was random. Conversely, a randomized experiment does not automatically represent every nonparticipating population. State the units, assignment mechanism and population limits before making either claim.