Tags: finite-sample/simcheck

simcheck 0.1.0

Monte Carlo tests for statistical estimators, with the tolerance derived from
the replicate count rather than picked by hand.

This release adds the two things the fleet had been working around by hand:
interval width, so a vacuous interval can no longer satisfy a coverage gate,
and power. It also derives the last hand-picked tolerance in the package --
assert_se_calibrated's 0.15 -- from the sampling distribution of the ratio,
estimating kurtosis from the study rather than assuming normality.

The PyPI publish job is expected to fail: no trusted publisher is registered
for finite-sample/simcheck yet. The build, twine check and GitHub release
succeed, and a workflow_dispatch against this tag can complete the upload
later without a version bump.