7.3 Dataset-level metadata

Information that applies to the complete file should be recorded once rather than repeated in every variable row. This commonly includes:

• dataset name and version;
• date received or extracted;
• source system and responsible owner;
• purpose for which the data were originally collected;
• number of rows and columns at the time of review;
• unit of analysis;
• primary and foreign keys;
• time and geographic coverage;
• expected update frequency;
• confidentiality or access classification; and
• lifecycle status, such as draft, current, superseded, or archived.

Dimensions and missing counts change when the dataset changes. These values are meaningful only when they are attached to a date or version.