Studio 6: Govern data and measurement
This studio is part of a book in active development and has not yet been through the author’s review. Content may change as the review advances.
What you can defend when you leave. Establish where your numbers came from and whether they measure what you say they measure.
The milestone ahead
Results inherit the quality of the data and the measures beneath them, and no later check can repair what goes wrong here. This milestone settles how data reach you, where every number came from, and whether your instruments measure what your contract says they measure. The work is deliberately unglamorous. It is also where more projects quietly fail than anywhere else in the book.
This studio closes with Milestone 6: Your data and measurement, governed, a short chapter of its own after the lessons. What it asks you to produce. Provenance documentation, a data-management record, your measurement specification, and a route-specific permission recheck.
Milestone 7 runs the declared analysis on these governed data and produces your first reproducible result. The milestone chapter keeps the working details: the practice steps, the versioned record, and how the artifact is assessed.
Before you can work this studio
This studio assumes data have reached you, and no earlier studio arranged that arrival. Settle acquisition first, in writing, before the steps below.
Your permission status from Studio 4 decides your route.
- Cleared or formal determination granted. Name your sampling frame, how you will reach those units, and what you will do about the ones you cannot reach. Non-response is a design fact, not an accident.
- Pending. Do everything that does not touch people: build and pilot the instrument on yourself, write the analysis code against simulated data of the shape you expect, and wait.
- No authority to ask, or not authorized. Take the no-permission route, which reaches every method in this book: published aggregate statistics, an openly licensed dataset with no individual records, or data you simulate yourself. Write down which you chose and what it costs you, because a simulated dataset can demonstrate a method but cannot support a claim about the world.
Whichever route you take, record where the data actually came from before you touch them. That record is the first row of the provenance documentation this studio produces.
The lessons in this studio
Read them in order, working each lesson’s It is your turn as you go. Each one builds a piece of Milestone 6; beside each lesson is the piece it hands you.
- Lesson 20 — Data Provenance and Data Quality: your acquisition route, written and dated, your provenance record for every dataset and borrowed number (the primary source behind your headline claim opened and read), your data-management update, and the permission recheck against the data as they actually arrived.
- Lesson 21 — Measurement and Operationalization: your measurement specification: concept, construct, indicator, the meaning sentence for each number, the reliability check’s result, the indicator-to-use validity argument with its strongest rival reading, and the boundary that follows.
When the last lesson closes, the milestone chapter is next. Nothing you write in a lesson is busywork: every piece above is on the milestone’s checklist.