Part II — The Data Chain

This is the production spine: standards-compliant datasets in, analysis-ready datasets out, with traceability preserved at every joint. The two chapters here are the most code-heavy in the book, and the most directly reusable in day-to-day study work.

Chapter 4 teaches the admiral LEGO method — building ADaM datasets from composable derivation bricks, where each brick is small enough to validate and the assembly reads like the spec. Chapter 5 closes the loop: your define spec becomes code, checks, labels, and transport files through metacore and xportr, so metadata stops being documentation about the data and becomes the data’s enforceable contract.

By the end of this part you should be able to take one dataset from your own shop’s spec and rebuild it as bricks with a metadata-driven export — that rebuild, not the reading, is the completion criterion.