7.5 Source, standardized, and derived variables

Preserve the original source field. Create new variables for corrections, standardization, and analysis. This makes the transformation visible and allows the analyst to compare the result with the original.

For example, NVRW’s source field home_site contains Central, CENTRAL, Ctr, and 01. The team creates home_centre_std using a documented mapping. A separate field, centre_region, groups centres for a regional comparison. The dictionary identifies the first field as source data, the second as standardized, and the third as derived.

Derived variables need enough detail to reproduce them. “Member tenure” is incomplete. A useful entry identifies the dates used, unit, treatment of partial periods, and reference date.

Worked Example: Writing Data-Dictionary Entries

Variable	Definition	Type and format	Allowed values	Missing rule	Source or derivation
member_id	Anonymous identifier for one member	Character, 10 positions	Unique nonblank code	Missing not permitted	Membership system
visit_time	Recorded facility check-in time	Date-time, local time	Valid time within extract period	Blank means entry was not recorded	Access system
contribution_margin	Revenue less directly attributable service or product cost	Decimal, dollars	May be negative after refund	Missing when direct cost unavailable	net_revenue - direct_cost
retained_12m	Whether a member has an active membership 12 months after the initial start	Logical	TRUE, FALSE	Missing when 12-month observation window is incomplete	Derived from membership periods

The entries reveal that retained_12m cannot be calculated for recent members. Treating those cases as non-retained would bias the analysis.