rdhte 0.2.0

First systematic audit pass following the 0.1.0 pre-release. Fourteen HIGH-severity and seven MED-severity bugs were fixed across the R and Stata code, plus a new data = argument and a multi-script verification harness. A 2026-05-01 Windows verification pass uncovered three further HIGH-severity bugs (Stata weights not forwarded to bandwidth selection; R rdhte_lincom silently applied multcomp’s single-step multiplicity adjustment; harness orchestrator regex/CSV-staging gaps) which are included below.

A subsequent expert-review pass on the Stata code on 2026-05-01 PM uncovered six further HIGH-severity issues, surfaced one previously silent latent inference bug in the no-covs_hte path, and bumped the Stata floor from 14 to 16 (frames). All are included in this version.

Stata 16+ now required (was Stata 14+). Driven by the migration from preserve to frames for the in-program working sample.

New features

Bug fixes (R)

Bug fixes (Stata)

Verification harness

Eight new top-level scripts under the project root provide a verification gate analogous to rdrobust’s:

run_rdhte_checks.py orchestrates all of the above end-to-end and reports per-spec |R-Stata|. Marker on success: RUN_RDHTE_CHECKS_OK.

Two orchestrator-level bugs found and fixed during the 2026-05-01 Windows verification pass:

Stata expert audit (2026-05-01 PM)

A second pass on the Stata code, severities HIGH / MEDIUM / LOW. All HIGH fixed; most MEDIUM applied; LOW partially done.

HIGH

MEDIUM

LOW

MEDIUM items deferred (not in this release)

rdhte 0.1.0

Initial release.