Skip to content

Detective Questions for Undocumented Data Sets

Prefer to fill it in?

A fillable version of this tool works right in your browser: open the interactive version. Add your own rows, columns, and items, then download, copy, or print your work. Nothing leaves your device.

When documentation is incomplete, analysts can approach the data like detectives.

Ask how the data might have been created

  • Was the data set built from surveys, clinical visits, or EHR data?
  • Were participants followed over time?

Look for study phases or waves

  • Do variable names suggest baseline or follow-up measurements?
  • Are repeated measurements present across time?

Identify domains

  • Do clusters of variables relate to imaging, genetics, medications, or other clinical areas?

Find likely stratification variables

  • Are there site, sex, mutation type, or age-group variables?
  • Are these complete and consistently coded?

Look for naming clues

  • Do prefixes or suffixes suggest questionnaires, modules, or scoring systems?

Inspect time patterns

  • Do some variables appear only after certain years?
  • Could that reflect a change in study design?

Watch for repeated records

  • Are multiple rows linked to the same participant across visits?

Check whether data types make sense

  • Are numeric variables stored as text?
  • Do categorical variables contain inconsistent values?