Module 6 — From Data to Auditable Evidence

Data Provenance

Trace sources, joins, definitions, exclusions and time ordering from raw records to analytic variables

Data Provenance

Every coefficient begins as a data lineage

For Pathways:

application registerlottery fileoffer logeducation linkageanalysis table.\text{application register} \rightarrow\text{lottery file} \rightarrow\text{offer log} \rightarrow\text{education linkage} \rightarrow\text{analysis table}.

If an applicant ID is duplicated during linkage, the standard error and treatment mean may both change. “Data cleaning” is not neutral when it changes who represents the estimand.

Keep four layers separate

LayerPurposeRule
source/rawimmutable acquired filesnever hand-edit
intermediatetyped, deduplicated, linked recordsone scripted transformation per object
analyticone row per declared unit/timevariables match the identification memo
outputtables, figures, model objectsgenerated, never manually corrected

Store checksums or source version identifiers. Record access dates because live administrative extracts and web APIs change.

A useful data manifest

FieldPathways example
source ownerNorthbridge scholarship office
file/versionlottery register, 2024-09-03 extract
unit and periodapplicant, 2024 cohort
keyapplication ID; unique before linkage
coveragesubmitted applications in participating districts
treatment timeoffer issued by 31 August 2024
outcome timedegree completion by 31 July 2029
restrictionsde-identified secure access only
transformationstatus codes mapped by versioned dictionary
known limitationappeals after 31 August require separate flag

This reveals a maturity problem immediately: five-year completion for the 2024 cohort is not observed in 2026. Use an earlier cohort, a shorter-horizon outcome or describe the analysis as prospective.

Joins need assertions

Before a merge, test key uniqueness and expected coverage. After it, report:

  • rows before and after;
  • unmatched keys on each side;
  • one-to-one, one-to-many or many-to-many relationship;
  • duplicated units created;
  • match rates by treatment and outcome-relevant groups.

An overall 95% match rate can hide 99% among controls and 85% among treated units.

Time-stamp variables by causal role

VariableMeasuredRole
prior scorebefore assignmentcandidate confounder/precision variable
offerassignment dateinstrument or treatment
scholarship receiptafter offertreatment uptake/mediator for ITT
first-year enrolmentafter offeroutcome or mediator for later completion
completionlateroutcome

The same field can be an outcome in one analysis and a mediator in another. A data dictionary should store definition, timing, coding, missing-value meaning and permitted role—not only column type.

Privacy and provenance reinforce each other

Keep direct identifiers outside the analysis repository. Document linkage without exposing keys, secrets or row-level personal data. For restricted sources, publish:

  • data-access route and eligibility;
  • non-disclosive metadata and variable construction;
  • synthetic or simulated fixtures that exercise the pipeline;
  • disclosure-control steps;
  • code that runs once authorised data are mounted.

Quick check

The treatment indicator is reconstructed from the amount paid during the year. Which estimand is threatened?

Answer
An offer ITT cannot use payment as assignment: payment is post-offer take-up. Preserve the lottery/offer field for ITT and use receipt only with an IV or another explicit receipt-effect design.

Next: Reproducibility and Reporting

Copyright © 2026