AHTauDesign 1.0
Major update
- Expanded AHTauDesign from truncation-time assessment to prospective
trial planning for average hazard estimands.
- Added null calibration and support-rule freezing.
- Added
simulate_design() as a public interface for
generating simulation results.
- Implemented Stage 1 support audits and conditional Stage 2 horizon
selection.
- Added independent evaluation of selected planning alternatives.
- Added Part II reverse design with coarse search, fine localization,
and independent validation.
- Added auditable planning reports.
API changes
- Replaced the earlier
assess_tau(),
optimize_tau(), and twostage_tau() workflow
with the new planning interfaces.
- Removed legacy IR outputs from Stage 2.
- Standardized the signed power comparison as
power_difference.
- Added input validation and expanded automated testing.
Important notes
- Information support and statistical power are evaluated
separately.
- A supported clinical horizon is retained.
- Alternative horizons are considered only after Stage 1 support
failure.
- Support-rule thresholds require setting-specific calibration or
validation.