On this page
Specify the population claim
A fictional study asks for average complaint-resolution time for all cases opened during a quarter. Its target contains cases, rather than customers or employees. A customer with five cases contributes five units if the question concerns cases, but a customer-level question would require a different construction.
Inspect the actual selection list
Suppose the analyst selects from a file of completed cases only. That frame omits cases still open at the extraction date. Random selection from the file does not bring those missing units into the sample. The analyst must change the frame, change the estimand or explicitly investigate the omitted coverage.
Match rows to units
| Layer | Definition | Potential mismatch |
|---|---|---|
| Target | Cases opened in the quarter | Unresolved cases still belong |
| Frame | Completed-case export | Open cases absent |
| Stored row | One status update | Several rows may refer to one case |
Write a definition that another analyst can apply
Record inclusion dates, identifiers, duplicate treatment and whether unresolved cases are retained. A reproducible unit definition is needed before counting n. The storage row count is not automatically the number of independently sampled units.