Changelog

All notable changes to this project are documented here.

The format follows Keep a Changelog. Set the release version with uv version X.Y.Z, record the release here, and tag the merged commit vX.Y.Z. Pushing the matching tag builds and publishes that version.

Unreleased

0.1.0 - 2026-10-15

Added

  • fsm and fsm_batch: the Finite Selection Model with all eleven selection functions of the R package FSM, reproducing its allocations exactly given the same selection order.

  • som (SCOMARS, global percentage, randomized chunk) and crd.

  • perm_test, ess_rand, ess_model, tasmd, tasmd_rand, make_sq_inter, asmd and love_plot.

  • The Lalonde (Dehejia–Wahba NSW) data, rebuilt from the NBER source.

  • R parity fixtures, the suite that checks them, and a weekly workflow that regenerates them from a fresh R install.

Fixed

  • Impute missing potential outcomes for general two-group nulls even at zero tau.

  • Apply the Monte Carlo p-value correction to sampled reference assignments; expose exact enumeration explicitly.

  • Reject unsupported rank-test nulls and invalid inference inputs, and clarify that multi-group tests use the sharp no-effect null.

  • Keep numerical tie handling relative to the observed statistic’s scale.