On this page
Define the target and units
Write the population, reference period and quantity to estimate. Check whether a row is a case, a person, a transaction or a repeated update. Identify any changes in population definition during the research.
Reconstruct the selection procedure
| Layer | Check |
|---|---|
| Frame | Coverage, duplicates and date |
| Randomization | Possible samples and selection probabilities |
| Groups | Strata or cluster stages and allocations |
| Response | Selected, responding and usable counts |
| Estimator | Population weights and target units |
| Uncertainty | Dependence and variance convention |
Perform an independent small check
When possible, enumerate a tiny fictional version of the design. Unit probabilities should reconcile with the expected sample count. Weights should match the intended target and sum as required. A full-population simple random sample should have zero subset-selection variance.
Review the conclusion’s scope
An estimate may be precise within the covered population while failing to cover an omitted group. A random selection procedure does not establish causality, and a count of thirty does not establish normality. Report the assumptions and remaining limitations next to the result.
Continue with CFA learning resources