Datums and probability maps | Lateral range curves and sweep width | Effort, coverage and probability of detection | Optimal search | Your own search | Planning a search, start to finish | On a grid of your own | References

1. Where the plume goes | Dispersion coefficients | Plume rise | Ground-level concentration | 2. From concentration to deaths | Concentration-response | Attributable fraction and avoided deaths | The burden chain | 3. Who bears it | 4. Estimating the exposure effect from your own data | 5. The whole pipeline, with an assumption log | 6. What did this computation cost? | Utilisation: measured, per process or per machine | The grid: the latest year, every country, offline | First Nations, Métis and Inuit populations | References

Attesting a manifest | Bundling a capsule | A chain of manifests | Sealing the chain | Which to use | Releases of Indigenous health tables | The table in a picture

Parsing band labels | Representative values | Expanding to per-unit values | Concentration | The assumption you cannot check: sensitivity to the open band | Heavy tails | Workflow | How unevenly a population is spread | The bands and the curve in a picture

1. The provenance record | 2. Resolving the live source (network) | 3. Synthetic fallback: the pipeline always runs | 4. Integrity and schema validation | 5. The manifest: account for everything | Where this sits in the MORIE family | A capsule for an Indigenous health table | The capsule's tables in a picture

The report's numbers against the table's labels | Codes that arrive without their labels | A recode you can prove | Counts against what was published | The record | The transfer that matters: SPSS (or Stata, SAS) to R | Exonerating the software, or not | The rotation in a picture

A known effect, recovered | The NSW job-training experiment, and what observational data can recover | Overlap and balance | Indigenous examples | References

Two ways to the variance | Estimates by domain | Comparing groups | The Canadian Substance Use Survey, against the published figures | When you only have the published table | Linearization or replicates, the same answer

Is the court keeping up? Case flow and clearance | How long do cases take? A life table from banded times | How do cases end? The mix of outcomes | Is a change in the clearance rate real? Simpson's paradox and case mix | Where do incidents cluster? Kernel density | Do incidents repeat nearby, soon after? The Knox test | Do the hot spots last? | Indigenous life expectancy | References

Difference in differences | Before any model | Leads and lags | Staggered adoption | Inference from the assignment | Counts | One treated place | Measures for justice data | Doing better than the original designs | Indigenous examples | Acknowledgements | References

Falsification controls | Power curves | Families of statistics: the multiple-testing correction | Sensitivity to an unmeasured confounder | Where this fits in a capsule | Several gaps, one family of tests | The tests in a picture | References

Whose data these are | Reading a table as published | Perceived mental health, against non-Indigenous adults | The whole scale, not one cut of it | Age standardisation | Wellness beside distress | Holistic balance | The four directions | Healing as a process | Connection across places | Community-set weights, and the Community Well-Being index | The 2021 Census | Intergenerational exposure | The Calls to Action | Cultural continuity | Your own data | What the package does not compute | Acknowledgements | References

A re-release that is statistically the same | A silently rescaled column | Looking at the drift | Which test, and why | Screening for figures that were not measured | Structure, separately | Two cycles of the same survey

Digests of text and files | Digests of R objects | Keyed digests and constant-time comparison | Deriving a key from a passphrase | Randomness | Merkle trees: pin a large file, verify a piece | How these implementations are checked | Pinning a community's copy of a table | The pinned table in a picture

Simulating a Hawkes process | Fitting | Checking the fit: time-rescaling residuals | Events recorded to a resolution | Evaluating a likelihood directly | Boundaries and identifiability | Indigenous examples | References

Encoding R objects | Parsing JSON | A repeated key is an error | Big integers | Exact serialisation: types and attributes included | Compressed, base64 JSON | Why the parser is held to a security standard | Sending a table as JSON | The table after the round trip in a picture

ML-KEM | Implicit rejection | Masked decapsulation (the default) | Key validation | HQC | Reproducible keys for tests | Using the shared secret | How these implementations are checked | Sending a table to a community data steward | The sizes in a picture

The route order | Asking, with the packages' help at hand | The signed services document | What the package will not do on the network | The data key | From the command line | Asking about the Indigenous tables | The providers in a picture

One command | Per column | Where the gaps are | Is the missingness itself a problem? | Rows that do not belong | Figures that were not measured | What none of this does | Describing an Indigenous health table | The distributions in a picture

Missingness | Duplicates | Multivariate outliers | Distinct counts in fixed memory | Does the data look fabricated or truncated? | Order of operations | Suppression in published tables | The outliers in a picture

The schema | Parsing a report | The pieces | Written for hostile input | Fetching live reports | Indigenous police services | The report's timeline in a picture

One call per result | Air pollution | Courts and crime | First Nations, Metis and Inuit health | Evaluating a policy | Customising

Why post-quantum, and why three schemes | ML-DSA: the default | Context strings bind a signature to its purpose | Hedged and deterministic signing | Signing a digest computed elsewhere | SLH-DSA: hash-based, stateless | Pre-hash variants | XMSS: hash-based, stateful | The key file is the state | Every way of being wrong returns FALSE | How these implementations are checked | Signing a release of Indigenous tables | The sizes in a picture

Signing: a shared secret | Signing: a public verifier, after quantum | One signature per index, and no more | Signing something a verifier can check | The number in the manifest is the number | Whether the finding survives | Re-deriving the numbers, not just checking them | Saying in advance what you will report | One file to hand someone | When, not just in what order | Pinning chunk by chunk | The history, not just the run | Sign the seal, not the head | What is verified, and what is not | Provenance for Indigenous data | The pinned data in a picture

1. Timing | The threat | Dynamic check: ctgrind (inst/ctcheck) | Measured on hardware: dudect (inst/dudect) | What this found | 2. Power and electromagnetic emanation | The countermeasure: first-order masking | Why the share-pair operations are assembly | Simulated power analysis: TVLA (inst/tvla) | 3. What is outside these checks | 4. Choosing the plain path | Indigenous examples | Constant time in a picture

Expected counts by indirect standardisation | Direct standardisation, for populations of different ages | Standardised ratios with exact intervals | Empirical Bayes shrinkage | Funnel plots | Spatial autocorrelation | From a published table to a judgement | References

A Statistics Canada table, as published | Bands | Concentration | Trend across a handful of years | Did it change when the policy changed? | Points, and the regions that contain them | Recomputing is not the same as checking | A second route can disagree | The coverage share is not a denominator | Regions | Autocorrelation, if you have a neighbour list | Indigenous examples | The results in a picture

Average daily population | Average length of stay | Admissions implied by stock and stay | Stock and flow together | The distribution of stays | Periods and fiscal years | From a published table | Indigenous examples | The figures in a picture | Reference

The page | Data | The tabs | The same work in a script

A monotone trend: Mann–Kendall and Sen's slope | The rate of change: a Poisson trend | A step change | Two survey periods | A pace, projected | Choosing between them | The trend in a picture | References

Rules | Ready-made rules | Inferring a schema | Synthetic stand-ins | Putting them together | Rules for a published table | The synthetic stand-in in a picture

The data | The first failure: matching on row order | The second failure: a percent off a small base | The third failure: a percent of a percent | Seasonal series | Across the span | Rates, when the denominator moves too | Output | Indigenous examples | The change in a picture

The table | One call | What the release withheld | Trend | Has the data changed since last time? | Stock and flow | Other published tables | Record it | The analysis in a picture

Write the record | Verify it | Record the results, sign and bundle | Check a fresh download for drift

1. Browsing the bundled store | 2. The named loaders | Chicago crime samples | Ontario SIU director's reports | CIHI data-table catalogue | 3. Data integrity | 4. Differential privacy | 5. Re-identification risk | Bridging to Python

Overview | Browsing the catalogue | Per-dataset detail | Loading data | Bundled tables | Statistics Canada / Health Canada data acknowledgment | Where to go next

What this vignette covers | Group-disparity metrics | Predictive-policing calibration audit | Further reading

Overview | A reproducible synthetic example | Single-robust ATE / ATT / ATC | Doubly robust: augmented IPW | Per-unit CATE (T-learner / S-learner) | Group-level GATE | Sensitivity analysis: the E-value | Where to go next

Overview | A 2x3 contingency table | Effect size: Cramer's V | One-way ANOVA + omega-squared | Published chi-square statistics (Sprott, Doob and Iftene) | Where to go next

Overview | The CPADS column contract | Loading and canonicalising | Survey-weighted proportions | Propensity-score IPW workflow | Loading the real PUMF | Where to go next

What this vignette is for | 1. How large must the sample be? | 2. Reaching the group: oversampling and screening | 3. Making the sample agree with known totals | 4. When the identifier is wrong | 5. Measurement: does the instrument mean the same thing in both groups? | 6. Governance is part of the method | Summary

Overview | Cohen's d (continuous, two-group) | Cramer's V (categorical association) | Omega-squared (one-way ANOVA) | Proportion confidence intervals | E-value (sensitivity analysis) | Where to go next

P4. Predictive-policing feedback loops run away, and slowly | P5. What a risk score's fairness can and cannot be | P1. The dark figure as an interval, not a point | P3. Hot-spot effects when places affect their neighbours | P7. How much of the "law of crime concentration" is crime-free places? | P2, P6, P8 in one breath | References

Overview | Loading the package | The dataset catalogue | Loading a dataset | A simple ATE estimate | Companion estimators | Where to go next

Overview | Setting up | Estimating propensities | Hajek-stabilised IPW | Weight diagnostics | AIPW for protection against IPW failure | Where to go next

Overview | Included reference samples | OTIS via CKAN (data.ontario.ca) | TPS via ArcGIS Open Data | SIU Director's Reports | Workflow: from included sample to verified result | Canadian case law and legislation via A2AJ and CanLII | Caching layout | References

Overview | OTIS suite | Placement-count concentration on b09 | Segregation-duration KM on b01 | Mortification co-occurrence (alert columns) | Region locality | Mandela classification | TPS suite | Levy-flight Hill exponent on inter-event step lengths | Moran's I + DBSCAN clustering | Neighbourhood inter-event recurrence | Hawkes manifest loader | SIU suite | References

What this vignette covers | Where the OTIS data come from | The canonical a01 design | The ten-estimator ensemble | The general multi-estimator entry point | Aggregate companion: incidence-rate ratios | Mandela classification | Federal counterparts | Where to go next

Difference-in-differences with automatic staggered handling | Instrumental variables with a weak-instrument gate | Regression discontinuity with bundled diagnostics | Propensity-score weighting | One pipeline end to end

Overview | A synthetic two-tone signal | Basic FFT | Spectral density estimates | Short-time Fourier transform | Envelope and cepstral pitch | Where to go next

General principle | Applied example

Overview | Stratified, cluster, and PPS sampling | Calibration weights and design weights | Effective sample size and design effect | Bootstrap variance with weights | Where to go next

The data | 0. Guard the categories first (module 25) | 1. The DAG (module 13) | 2. Identification | 3. Matching on the identified set (module 1) | 4. Estimation through the DAG (modules 10 + 13) | 5. Refutation (module 13) | 6. The MRM pipeline and a publication table (module 24) | Where to look next

Overview | Inspecting the shipped manifest | The corpus | Auditing a single case | Comparing parser output against an external table | LLM extraction | French -> English translation | Corrections: the parser learns | Format validity | Where to go next