Overview
Open the primary tutorial link
- Overview
-
Nine public chapters on building a research database that can say how much it should be trusted, from designing the first table to publishing a citable dataset. The lab’s Taiwan Politics database is the worked example.
Content
1. What a Research Database Is For
A research database preserves evidence along with an account of its limits. This chapter introduces the admit-then-audit approach and the eight-stage lifecycle.
2. Designing the Schema
A useful first schema has to leave room for discoveries during research. This chapter explains stable identifiers, provenance hooks, and the handoff from dbdiagram to Django migrations.
3. Raw Data, Staging, and Loading
Staging makes a source file loadable while retaining its original form. This chapter follows a spreadsheet through mapping, loading, immutable raw capture, and verification.
4. Provenance
A source and a locator identify where a claim came from. This chapter separates evidence about a record from evidence about a violation and shows why both are needed.
5. Intake and the Legacy-Field Pattern
Legacy fields retain what the source said while validated fields record what the lab can establish. This chapter introduces the six-value status vocabulary and the difference between bootstrap and ongoing delivery.
6. Rules, Versions, and Bindings
An audit rule applies a specific standard to specific records at a specific time. This chapter explains rule types, definitions, versions, bindings, and priority segments.
7. One Rule, End to End
A rule takes more than working code to reach a researcher. This chapter follows one rule through its manifest, checksums, rehearsal, and production gates.
8. Identity Resolution
Identity decisions determine whether records refer to one person or several. This chapter covers minting, adjudication, consolidation, blocking, and the risk of transitive matches.
9. Derivation, Release, and Download
Derived data is computed from collected records and must have a refresh contract. This chapter explains versioned releases, generated codebooks, disclosure, and citation.