- Published
- Updated
Replacing blank fields with zero may make a dataset technically complete while changing its meaning. An unperformed measurement, an unknown result and a genuine measured zero represent different states and should not be conflated.
Preserve the meaning of data
Define how reasons for missingness are represented during collection and transfer. Putting a special numeric code into a result field can cause calculations to treat it as an actual measurement. Link the code list, field definition and analysis use explicitly. Also define how a result obtained later updates the earlier missingness status without concealing its history.
Worked example
For example, the source marks an assessment as not performed with an explanation, but a transfer fills the empty numeric field with zero. A summary report then displays an unusually low result. The team transfers the missingness reason separately and reviews historical zeros against source records. Genuine zero measurements are not automatically reclassified as missing.
Review checks
- Test blank, zero, not-done and unknown examples separately.
- Confirm that missingness codes do not enter numeric summary calculations.
- Preserve the relationship between later results and earlier missingness reasons.
Acceptance and boundaries
Acceptance means a reviewer can distinguish an observation from a placeholder. Selecting missingness codes does not by itself determine the statistical treatment of missing data. Appropriate experts should document analysis decisions separately. The operational aim is explicit, reusable meaning rather than a misleading appearance of completeness; a screen containing no empty fields is not necessarily a high-quality record.
Frequently asked questions
- Is one missingness code enough?
- The intended use may require distinctions between reasons, although unnecessary complexity can increase user error.
- Can zero ever be used?
- Yes, when it is the actual observation. The problem is presenting an unobserved value as zero without explanation.