On this page
Identify the first-stage group
A fictional audit groups records into eight equal-sized branches. The branch is the first-stage selection unit. Selecting a branch and selecting an individual record are different random events and should have separate entries in the plan.
State what happens inside a selected group
| Design | First stage | Inside selected clusters |
|---|---|---|
| One-stage | Randomly select clusters | Observe every unit |
| Two-stage | Randomly select clusters | Randomly select some units |
Follow both stages for a unit probability
If two of eight branches are selected uniformly and ten of fifty records are selected uniformly within each chosen branch, a particular record has inclusion probability (2/8) × (10/50) = 0.05. This relies on the explicitly stated equal group sizes and selection rules.
Read the design’s uncertainty separately from its cost
Records from the same branch may be similar. Treating all collected records as independent can miss that structure. Additional stages can reduce collection effort, but their precision depends on allocations and dependence. There is no universal ranking of error based only on the number of stages.