What is and is not in this version | Five ideas worth knowing | 1. Everything carries its provenance | 2. Citations come from the data, not from you | 3. Access conditions are data, not decoration | 4. Failure is explicit | 5. Everything is a tibble that speaks tidyverse | A first session | Configuration | Caching | Where to next

The reference | The two classic corruptions | A worked join | Gross County Product | Health by county: the KDHS | Matching county names across sources | What is deliberately absent

Registered sources | What each stability level means | What works | What does not, and why | The archive's microdata cannot be downloaded | Documentation files are not listed | The newest Gross County Product years are still documents | Geography: one vintage, no boundaries, no classifications | Variable documentation is uneven | How you will know if something changes | Reporting a change | Disclaimer

What is supported | Finding a survey | Reading a study's documentation | Searching variables | Files, and what you may do with them | Downloading microdata | Open microdata: the FinAccess surveys | Handling errors | Citing what you used | Caching

Curated, not guessed | Listing and fetching | Reading the columns | County tables join everything else | Units, footnotes and sources | Into the tidyverse | Provenance travels | What this is not

The Data Portal (SDMX) | Open Data for Africa (the republisher) | Consumer prices | CPI geography is not census geography | Retail prices | Census tables | What the census source is, and is not | Finding a table | Getting observations | Filtering geography | A worked example | Provenance and citation | Honesty notes

Events, reporting coverage and exposure | Missing ethnicity changes the estimand | Denominator scenarios require comparable cells | Direct age-sex standardisation | Rankings need uncertainty, too | Validation and limits

Acquire a snapshot, then choose whole files | Coverage comes before comparison | Attach geography locally | Inspect timestamp quality and exposures | Export a reviewable report

Three separate meanings of an outcome | A time-quality screen is mandatory | Local clock time, sunset and civil twilight

A spatial model of disparity | Recovery independent of total intensity | Replicated evidence, including failure modes

The canonical model | Solving and checking | Monetary transmission | History and forecast | Specification checks | Adapting the model to your country | Real data: Czechia | Policy analysis: conditions, judgment, rounds | Closing the round: audit and report | Where to go next

Priors | A laboratory: estimating an AR(1) | Identification: ask before you sample | Marginal likelihood and Bayes factors | On real data

Events, reporting coverage and exposure | Missing ethnicity changes the estimand | Denominator scenarios require comparable cells | Direct age-sex standardisation | Rankings need uncertainty, too | Validation and limits

Acquire a snapshot, then choose whole files | Coverage comes before comparison | Attach geography locally | Inspect timestamp quality and exposures | Export a reviewable report

Three separate meanings of an outcome | A time-quality screen is mandatory | Local clock time, sunset and civil twilight

A spatial model of disparity | Recovery independent of total intensity | Replicated evidence, including failure modes