Release notes¶

This project follows Semantic Versioning.

Unreleased¶

Renamed the distribution, import package, command, documentation, and repository from geoinference to geoestimate.
Replaced the proportion-specific entry point with an immutable Sample object and scalar Estimate results.
Added inference for means, population totals with known population size, ratios of totals, and means of row-level ratios.
Simplified uncertainty selection to design-based, iid, cluster-sandwich, or pairs-bootstrap inference with matching confidence intervals.
Reorganized the documentation around tasks, estimands, sampling assumptions, file workflows, and stable versus experimental APIs.

Require users to define the analysis domain before estimating row-level ratios with zero denominators.
Require a known row-level population size for population totals and state that finite population corrections are not implemented.
Validate every estimand’s variables and option domains at the public boundary.
Cross-check equal-probability analytic standard errors against the R survey package under equivalent designs.

Reject invalid counts, undefined samples, incomplete clusters, invalid confidence levels, and unavailable uncertainty methods at the public boundary.
Keep simulated visits inside their assigned shifts and reject infeasible routes.
Match convenience-scheduler capacity to its requested itinerary count.
Use common population draws when comparing simulation pipelines and methods.
Use the requested standard error when measuring simulation coverage.
Retain every bootstrap resample for which the estimand is defined.
Center two-sided Moran permutation tests on their permutation null distribution.

PPS, GRTS, annotation-subsampling, finite-population, and walk-spacing options that did not affect the estimator.
The unverified public wild-cluster-bootstrap interval.
Spatial diagnostics from the stable estimate result.