On this page
Inspect the frame against the target
A quarterly business survey may target all active firms but use a list that omits newly formed firms. Those absent units have no chance of selection from that list. Increasing the sample drawn from the old list does not give the missing firms an inclusion chance.
Track what happens after selection
Among selected firms, some may not respond or may omit a key variable. The number originally selected and the number contributing to an estimate differ. If response is related to the variable of interest, the observed responders need not behave like the original selected sample.
Keep counts for each stage
| Stage | Example count | Question |
|---|---|---|
| Eligible frame units | 1,000 | Who is missing from this frame? |
| Selected units | 100 | How were they selected? |
| Responding units | 70 | Why did 30 not respond? |
| Usable outcome records | 65 | Which five lack the required field? |
Evaluate adjustments rather than assuming a cure
Weighting or imputation can address specified problems under explicit assumptions. They do not automatically reconstruct outcomes for every missing unit. Report what is observed, what is modeled and which target population the resulting estimate can support.