Address PR #372 R7 review (1 P1, 2 P3): HAD methodology accuracy + toctree

P1 - HAD descriptions disagreed with REGISTRY.md and shipped code
- Reviewer caught factual errors I introduced when drafting HAD copy:
  - "Strictly positive heterogeneous dose" excludes Design 1' (the QUG
    case where d̲ = 0 is allowed; mass-point fraction at d̲ matters
    for the support-plausibility heuristic but `d.min() == 0` is the
    canonical Design 1' invariant per REGISTRY L2504-2508).
  - "HC2 / Bell-McCaffrey small-sample SEs" is wrong for HAD: the
    mass-point path only exposes `classical` / `hc1` (HC2 / HC2_BM
    raise NotImplementedError pending 2SLS-specific leverage); the
    continuous paths use the CCT-2014 weighted-robust SE from the
    in-house lprobust port.
  - "Per-horizon WAS effects with correlated standard errors and sup-t
    bands" is wrong: per-horizon CIs are pointwise by default and have
    NO cross-horizon covariance (deferred to TODO follow-up). Sup-t
    is available ONLY on the weighted event-study path via
    `cband=True`.
  - "Multi-period event-study extension" silent on staggered handling:
    the estimator auto-filters staggered-timing panels to the
    last-treatment cohort (which retains never-treated units as
    comparisons) per paper Appendix B.2 prescription.
- Rewrite docs/api/had.rst, README.md catalog one-liner, and
  diff_diff/guides/llms.txt catalog one-liner to match REGISTRY exactly:
  Design 1' (d̲=0 / QUG) and Design 1 (d̲>0) both supported; three SE
  regimes documented (unweighted CCT-2014; `weights=` shortcut with
  classical/hc1 mass-point; `survey=` with Binder TSL); pointwise CIs
  per horizon; sup-t restricted to weighted event-study + cband=True;
  last-treatment-cohort restriction noted explicitly.

P3 - docs/api/had.rst cited Calonico/Cattaneo/Titiunik (2014) but it
was not in docs/references.rst (CONTRIBUTING.md now treats references.rst
as the bibliography source of truth)
- Drop the CCT 2014 inline citation from had.rst; the bias-correction
  primitive is an implementation detail tracked in REGISTRY.md and not
  needed at the user-facing API page level. The CCF 2018/2019 nprobust
  port reference is preserved as "ported from the nprobust machinery"
  without a specific citation, also matching how other API pages handle
  underlying numerical-method machinery.

P3 - had.rst was added to docs/api/index.rst autosummary but missed
the Module Documentation > Estimators toctree at the same file
- Add `had` between `continuous_did` and `efficient_did` in the
  Estimators toctree so the new page is navigable from the RTD API
  landing page.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: igerber

README.md

@@ -110,7 +110,7 @@ Full guide: `diff_diff.get_llm_guide("practitioner")`.
 - [SyntheticDiD](https://diff-diff.readthedocs.io/en/stable/api/estimators.html) - Synthetic DiD combining standard DiD and synthetic control for few treated units
 - [TripleDifference](https://diff-diff.readthedocs.io/en/stable/api/triple_diff.html) - triple difference (DDD) estimator for designs requiring two criteria for treatment eligibility
 - [ContinuousDiD](https://diff-diff.readthedocs.io/en/stable/api/continuous_did.html) - Callaway, Goodman-Bacon & Sant'Anna (2024) continuous treatment DiD with dose-response curves
-- [HeterogeneousAdoptionDiD](https://diff-diff.readthedocs.io/en/stable/api/had.html) - de Chaisemartin, Ciccia, D'Haultfœuille & Knau (2026) for **no-untreated-unit panels** where every unit receives a strictly positive heterogeneous dose; local-linear estimator targeting the Weighted Average Slope (WAS) at the support boundary, with multi-period event-study extension. Alias `HAD`.
+- [HeterogeneousAdoptionDiD](https://diff-diff.readthedocs.io/en/stable/api/had.html) - de Chaisemartin, Ciccia, D'Haultfœuille & Knau (2026) for designs where **no unit remains untreated**; local-linear estimator targeting the Weighted Average Slope (WAS) at the dose support boundary (Design 1' with `d̲ = 0` and Design 1 with `d̲ > 0`), with a multi-period event-study extension (last-treatment cohort, pointwise CIs). Alias `HAD`.
 - [StackedDiD](https://diff-diff.readthedocs.io/en/stable/api/stacked_did.html) - Wing, Freedman & Hollingsworth (2024) stacked DiD with Q-weights and sub-experiments
 - [EfficientDiD](https://diff-diff.readthedocs.io/en/stable/api/efficient_did.html) - Chen, Sant'Anna & Xie (2025) efficient DiD with optimal weighting for tighter SEs
 - [TROP](https://diff-diff.readthedocs.io/en/stable/api/trop.html) - Triply Robust Panel estimator (Athey et al. 2025) with nuclear norm factor adjustment

diff_diff/guides/llms.txt

@@ -61,7 +61,7 @@ Full practitioner guide: call `diff_diff.get_llm_guide("practitioner")`
 - [SyntheticDiD](https://diff-diff.readthedocs.io/en/stable/api/estimators.html): Synthetic DiD combining standard DiD and synthetic control methods for few treated units
 - [TripleDifference](https://diff-diff.readthedocs.io/en/stable/api/triple_diff.html): Triple difference (DDD) estimator for designs requiring two criteria for treatment eligibility
 - [ContinuousDiD](https://diff-diff.readthedocs.io/en/stable/api/continuous_did.html): Callaway, Goodman-Bacon & Sant'Anna (2024) continuous treatment DiD with dose-response curves
-- [HeterogeneousAdoptionDiD](https://diff-diff.readthedocs.io/en/stable/api/had.html): de Chaisemartin, Ciccia, D'Haultfœuille & Knau (2026) for designs where **no unit remains untreated** (heterogeneous adoption / HAD); local-linear estimator targeting Weighted Average Slope (WAS) at the support boundary, with multi-period event-study extension. Alias `HAD`.
+- [HeterogeneousAdoptionDiD](https://diff-diff.readthedocs.io/en/stable/api/had.html): de Chaisemartin, Ciccia, D'Haultfœuille & Knau (2026) for designs where **no unit remains untreated**; local-linear estimator targeting Weighted Average Slope (WAS) at the dose support boundary (Design 1' with `d̲=0` and Design 1 with `d̲>0`), with multi-period event-study extension (last-treatment cohort, pointwise CIs). Alias `HAD`.
 - [StackedDiD](https://diff-diff.readthedocs.io/en/stable/api/stacked_did.html): Wing, Freedman & Hollingsworth (2024) stacked DiD with Q-weights and sub-experiments
 - [EfficientDiD](https://diff-diff.readthedocs.io/en/stable/api/efficient_did.html): Chen, Sant'Anna & Xie (2025) efficient DiD with optimal weighting for tighter SEs
 - [TROP](https://diff-diff.readthedocs.io/en/stable/api/trop.html): Triply Robust Panel estimator (Athey et al. 2025) with nuclear norm factor adjustment

docs/api/had.rst

@@ -1,33 +1,55 @@
 Heterogeneous Adoption Difference-in-Differences
 ================================================
 
-Estimator for designs where **no unit remains untreated** at the post period
-- every unit `g` receives a strictly positive heterogeneous dose `D_{g,2} > 0`,
-and there is no genuinely untreated control group to anchor a standard DiD
-contrast.
+Estimator for designs where **no unit remains untreated** at the post period.
+Every unit `g` is exposed to treatment at the same single date but adoption
+intensity (dose) varies across units; there is no genuinely untreated control
+group to anchor a standard DiD contrast.
 
 This module implements the methodology from de Chaisemartin, Ciccia,
 D'Haultfœuille & Knau (2026), "Difference-in-Differences Estimators When No
 Unit Remains Untreated" (arXiv:2405.04465v6), which:
 
 1. **Targets the Weighted Average Slope (WAS)** as the identified parameter
    when no untreated comparison group exists (paper Equation 2).
-2. **Uses local-linear regression at the support boundary** to estimate the
-   slope of the dose-outcome relationship at the lowest observed dose.
-3. **Provides bias-corrected confidence intervals** (Calonico, Cattaneo & Titiunik
-   2014 / `nprobust`-style) and HC2 / Bell-McCaffrey small-sample SEs.
+2. **Estimates WAS via local-linear regression at the dose support boundary**
+   for both Design 1' (the QUG / Quasi-Untreated-Group case where the support
+   infimum ``d̲ = 0``) and Design 1 (no QUG, ``d̲ > 0``).
+3. **Provides bias-corrected confidence intervals** ported from the
+   ``nprobust`` machinery for the continuous-dose paths, and a
+   structural-residual 2SLS sandwich for the mass-point path.
 4. **Extends to multi-period event-study settings** (paper Appendix B.2),
-   producing per-horizon WAS effects with correlated standard errors and
-   sup-t bands.
+   restricting staggered-timing panels to the last-treatment cohort (which
+   retains never-treated units as comparisons) with pointwise per-horizon CIs.
 
 .. note::
 
    **When to use HAD.** Use ``HeterogeneousAdoptionDiD`` when your panel has
-   no untreated unit at any treatment period (e.g. universal-rollout policies,
-   industry-wide tariff changes) but treatment intensity varies across units.
-   For panels with a never-treated control group and continuous treatment,
-   use :class:`~diff_diff.ContinuousDiD` instead. For binary reversible
-   treatments, use :class:`~diff_diff.ChaisemartinDHaultfoeuille`.
+   no untreated unit at the post period (e.g. universal-rollout policies,
+   industry-wide tariff changes) but treatment intensity varies across
+   units. For panels with a never-treated control group and continuous
+   treatment, use :class:`~diff_diff.ContinuousDiD` instead. For binary
+   reversible treatments, use :class:`~diff_diff.ChaisemartinDHaultfoeuille`.
+
+.. note::
+
+   **Inference contract.** Per-horizon CIs are always pointwise. There are
+   three SE regimes selected by call site:
+
+   - **Unweighted** - continuous paths use the CCT-2014 weighted-robust SE
+     from the in-house ``lprobust`` port; the mass-point path uses a
+     structural-residual 2SLS sandwich. No cross-horizon covariance.
+   - **``weights=`` shortcut** - continuous paths reuse the CCT-2014 SE;
+     the mass-point path uses an analytical weighted 2SLS sandwich
+     (``classical`` / ``hc1`` only - ``hc2`` / ``hc2_bm`` raise
+     ``NotImplementedError`` pending a 2SLS-specific leverage derivation).
+   - **``survey=``** - both paths compose Binder (1983) Taylor-series
+     linearization with ``df_survey`` threaded into ``safe_inference``.
+
+   A simultaneous confidence band (sup-t) is available only on the
+   **weighted event-study path** via ``cband=True``. Joint cross-horizon
+   analytical covariance is not computed in this release; tracked in
+   ``TODO.md``.
 
 HeterogeneousAdoptionDiD
 ------------------------

docs/api/index.rst

@@ -240,6 +240,7 @@ Estimators
    triple_diff
    trop
    continuous_did
+   had
    efficient_did
    two_stage
    wooldridge_etwfe