pretest-did 0.1.0__tar.gz

pretest_did-0.1.0/CHANGELOG.md

@@ -0,0 +1,59 @@
+# Changelog
+
+All notable changes to `pretest` will be documented in this file.
+
+Format: [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
+Versioning: [Semantic Versioning](https://semver.org/spec/v2.0.0.html)
+
+## [0.1.0] - 2025-12-22
+
+### Added
+
+**Core functionality**
+
+- DID estimators for δ̂ₜ and ν̂ₜ (Section 2.1)
+- Severity measurement with *p*-norm, *p* ∈ [1, ∞] (Section 3.1)
+- κ constant with closed-form solutions (Section 3.2, Proposition 1)
+- ψ function for Monte Carlo critical values (Section 5.1)
+- Conditionally valid confidence intervals (Section 5.1, Theorem 2)
+
+**Stata interface**
+
+- Complete command syntax with required and optional arguments
+- Formatted output display following Stata conventions
+- Comprehensive return values in `e()`
+- Event study visualization
+- SMCL help documentation (`help pretest`)
+
+**Python package**
+
+- Releasable `pretest` package with a typed public API and console script
+- Stata-style command parser, dataset validation, severity, kappa, confidence
+  interval, and high-level `compute_pretest_snapshot(...)` helpers
+- Monte Carlo simulation APIs including `compute_psi(...)`,
+  `compute_critical_value(...)`, `simulate_coverage(...)`, and
+  `simulate_coverage_from_covariance(...)`
+- Public Prop99 records-to-snapshot handoff plus local fixed-reference CLI
+  helpers for inspecting bounded same-case comparison metadata
+- Prop99 overall reference metadata now keeps `capture_ready: true`,
+  `graph-exported`, and `numeric_fields_promoted` attached to the captured
+  same-case metadata. The bundled Python split documents are record-backed:
+  deterministic estimator fields match the companion Stata capture, while
+  conditional interval bounds remain stream-specific because the Monte Carlo
+  critical value is computed by the Python RNG stream.
+
+**Two violation modes**
+
+- Iterative (default): CI = δ̄̂ ± {κ · Ŝₚᵣₑ + f(α, Σ̂) / √n}
+- Overall (Appendix C): CI = δ̄̂ ± {Ŝᐩₚᵣₑ + fᐩ(α, Σ̂ᐩ) / √n}
+
+**Additional features**
+
+- Stata command support for cluster-robust standard errors; the Python records
+  helper does not implement cluster covariance
+- Monte Carlo critical-value simulation support via the `simulate()` option
+
+### Requirements
+
+- Stata 17.0 or later
+- Python 3.11 or later for the `pretest` package