Distinguish Empty Survey Cells From Observed Zero Outcomes

Define the expected reporting grid

Before calculating subgroup results, list all planned combinations of the reporting variables. A missing output row can mean no sampled observations, no nonmissing outcomes, an excluded category or an unrequested factor level. These states should not be collapsed into a single numeric zero.

The survey grouped-statistics documentation includes controls for empty groups and missingness. Software defaults may drop an empty cell from a table, so create an explicit reporting grid and reconcile the output against it before publication.

Attach a reason to every absent estimate

For each cell, retain the unweighted count, number with an observed outcome, relevant weighted total and the status of the estimate. Use distinct labels such as no sampled observations, all outcomes missing, not eligible and suppressed under the disclosure rule. Preserve the reason in machine-readable output as well as the displayed footnote.

A model-based prediction for an empty cell is a different object from a direct survey estimate. If such a prediction is justified, label it and describe the model that supplies information. Do not allow a plotting routine to connect an absent estimate to zero and create an artificial downward trend.

Compare hypothetical cells

Suppose a regional table has three cells. Region A contains 20 observed people and no positive outcomes. Region B contains no sampled people. Region C contains 15 people, but all outcome values are missing.

For A, the observed numerator is zero and the denominator is known; a design-appropriate interval still needs consideration. For B, neither a direct sample proportion nor its usual interval is available. For C, participant presence does not supply an observed outcome denominator. Writing 0% in all three rows would imply three results that the data do not support.

Review sparse groups without changing the question silently

If categories are combined, document the scientific rationale, the timing of the decision and the new target definition. Avoid choosing combinations after inspecting which grouping yields a desired comparison. Retain the original grid in the review record so the change can be traced.

Check the survey provider’s reliability and disclosure policies before releasing small cells. Suppression should remain distinguishable from lack of data, while complementary cells may also require protection. In the final table, make the meaning of every symbol clear and ensure denominators are consistent across rows. A complete reporting grid should expose where evidence is absent rather than make the absence disappear through formatting.

Source verification: 6 November 2026. Examples are hypothetical.