Add yatchew_hr_test(null="mean_independence") mode

Mirrors R YatchewTest::yatchew_test(order=0). Closes the placebo
Yatchew R-parity gap from PR #392.

- New keyword-only `null: Literal["linearity", "mean_independence"]`
  on `yatchew_hr_test` (default `"linearity"` is bit-exact backcompat).
- `"mean_independence"` fits intercept-only OLS (residuals = dy - mean(dy));
  the downstream sigma2_diff / sigma2_W / sort-by-d machinery is shared.
- Wired through both unweighted and survey-weighted code paths
  (4-arm dispatch on (null × weighted)).
- `YatchewTestResults` gained `null_form: str = "linearity"` field;
  `summary()` renders the correct null-hypothesis title; `__repr__`
  and `to_dict()` updated.
- `tests/test_did_had_parity.py::TestYatchewParity` removed the
  placebo skip; routes effect rows through `null="linearity"` (R order=1)
  and placebo rows through `null="mean_independence"` (R order=0); both
  modes share the documented `× G/(G-1)` finite-sample convention shift
  and parity holds at `atol=1e-10`.
- New `TestYatchewHRTestMeanIndependence` class (15 tests) covering
  happy path, naive Python baseline at `atol=1e-12`, population-variance
  closed form, invalid value, default backcompat, mode-agnostic
  tie/constant-d rejection, NaN handling, weighted reduction at
  w=ones(G) at `atol=1e-14`, weighted non-uniform baseline,
  default-under-weights, survey×null orthogonality, the (linearity,
  weighted) baseline (4-arm coverage), zero/replicate-weight rejection,
  and G<3 mode-agnostic. One additive backcompat case in each of
  `TestYatchewHRTest` and `TestYatchewHRTestSurvey`.
- REGISTRY.md HAD § Yatchew note: TODO marker replaced with shipped
  description. CHANGELOG.md and TODO.md updated.

Patch-level (additive keyword-only kwarg + additive dataclass field
with default).

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

Author: igerber

CHANGELOG.md

@@ -9,8 +9,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 - **`ChaisemartinDHaultfoeuille.by_path` is now compatible with `trends_linear` (DID^{fd} group-specific linear trends) and `trends_nonparam` (state-set trends).** For `trends_linear`, the first-differencing transform runs once globally before path enumeration, so per-path raw second-differences `DID^{fd}_{path, l}` surface on `path_effects[path]["horizons"][l]` automatically. Per-path **cumulated level effects** `delta_{path, l} = sum_{l'=1..l} DID^{fd}_{path, l'}` (the quantity R returns under `did_multiplegt_dyn(..., by_path, trends_lin)`) surface on the new `results.path_cumulated_event_study[path][l]` field, mirroring the global `linear_trends_effects` cumulation. `to_dataframe(level="by_path")` exposes `cumulated_effect` / `cumulated_se` columns (always present, NaN-when-None — mirrors the `cband_*` convention from PR #374); `summary()` renders a "Cumulated Level Effects (DID^{fd}, trends_linear)" sub-section under each per-path block. SE on the cumulated layer is the conservative upper bound (sum of per-horizon component SEs, NaN-consistent), matching the global `linear_trends_effects` convention. Path enumeration runs on the post-first-differenced `N_mat_fd`: switchers with `F_g==2` fail the window-eligibility check and are dropped from path enumeration entirely (the existing global `F_g >= 3` warning still surfaces the issue), so a path whose switchers all have `F_g < 3` is silently absent from `path_effects` rather than present-with-NaN. Placebo under `trends_linear` returns RAW per-horizon values — there is no per-path placebo cumulation surface in either Python or R. For `trends_nonparam`, the set membership column is validated and stored once globally as `set_ids_arr`; the `set_ids` parameter is now threaded through the four per-path IF helpers (`_compute_path_effects`, `_compute_path_placebos`, `_collect_path_bootstrap_inputs`, `_collect_path_placebo_bootstrap_inputs`) so per-path analytical SE, bootstrap, placebos, and sup-t bands all consume the set-restricted control pool automatically. Per-period effects remain unadjusted under both extensions, consistent with the existing per-period DID contract. Validated against R via two new golden-value scenarios: `single_baseline_multi_path_by_path_trends_lin` (n_periods=13, F_g >= 4, cohort-single-path; per-path cumulated point estimates match R bit-exactly with `POINT_RTOL=1e-9`, cumulated SE within `CUM_SE_RTOL=0.20`) and `multi_path_reversible_by_path_trends_nonparam` (per-path point estimates AND placebos match R bit-exactly with `POINT_RTOL=1e-9`, per-path SE within `SE_RTOL=0.15`). **F_g=3 boundary-case divergence (`by_path + trends_linear`):** `F_g=3` switchers have only 1 valid pre-window Z value after first-differencing, triggering 30%+ relative divergence between Python and R per-path point estimates on paths whose switchers include `F_g=3`. A targeted `UserWarning` fires at fit-time on this regime; R parity is asserted only on the `F_g >= 4` parity fixture. Placebo parity for `trends_linear` is intentionally skipped (R's per-path placebo computation re-runs on the path-restricted subsample with different control eligibility than Python's global-then-disaggregate architecture surfaces; placebo + `trends_linear` is exercised via internal regression only). Cross-path cohort-sharing SE deviation from R documented for `path_effects` is inherited unchanged. Gates at `chaisemartin_dhaultfoeuille.py:1014-1023` removed; `by_path` docstring updated to add the two new compatibility paragraphs and remove `trends_linear` / `trends_nonparam` from the incompatible list. R-parity tests at `tests/test_chaisemartin_dhaultfoeuille_parity.py::TestDCDHDynRParityByPathTrendsLinear` and `::TestDCDHDynRParityByPathTrendsNonparam`; cross-surface regressions at `tests/test_chaisemartin_dhaultfoeuille.py::TestByPathTrendsLinear` and `::TestByPathTrendsNonparam`. See `docs/methodology/REGISTRY.md` §ChaisemartinDHaultfoeuille `Note (Phase 3 by_path ...)` → "Per-path linear-trends DID^{fd}" and "Per-path state-set trends" for the full contract.
+- **`yatchew_hr_test(null="mean_independence")` mode** mirroring R `YatchewTest::yatchew_test(order=0)`. Adds a `null: Literal["linearity", "mean_independence"]` keyword-only kwarg to `yatchew_hr_test`. Default `"linearity"` is bit-exact backcompat (residuals from OLS `dy = a + b·d + eps`, paper Assumption 8 / Theorem 7). New `"mean_independence"` fits intercept-only OLS (`dy = a + eps`, residuals `= dy - mean(dy)`); the downstream `sigma2_diff` / `sigma2_W` / sort-by-`d` machinery is identical between the two modes. Exposed on both unweighted and survey-weighted code paths (`weights=` / `survey_design=` compose orthogonally with `null=`). Adds a `null_form: str` field to `YatchewTestResults` so `summary()` renders the correct null-hypothesis description; `__repr__` and `to_dict()` updated. Closes the placebo Yatchew R-parity gap from PR #392 — `tests/test_did_had_parity.py::TestYatchewParity` now routes effect rows through `null="linearity"` (R `order=1`) and placebo rows through `null="mean_independence"` (R `order=0`); both modes share the documented `× G/(G-1)` finite-sample convention shift and parity holds at `atol=1e-10`. Patch-level (additive keyword-only kwarg + additive dataclass field with default).
 - **HAD `trends_lin=True` linear-trend detrending mode** on `HeterogeneousAdoptionDiD.fit(aggregate="event_study")`, `joint_pretrends_test`, and `joint_homogeneity_test`. Mirrors R `DIDHAD::did_had(..., trends_lin=TRUE)` (paper Eq. 17 / Eq. 18 / page 32 joint-Stute homogeneity-with-trends). Per-group linear-trend slope estimated as `Y[g, F-1] - Y[g, F-2]` and applied as `(t - base) × slope` adjustment to per-event-time outcome evolutions. Requires F ≥ 3 (panel must contain F-2). The "consumed" placebo at our event-time `e=-2` is auto-dropped (R reduces max placebo lag by 1 with the same effect). Mutually exclusive with survey weighting (`survey_design` / `survey` / `weights`): raises `NotImplementedError` per `feedback_per_method_survey_element_contract` (weighted slope estimator not derived from paper; tracked in TODO.md as a follow-up). Bit-exact backcompat for `trends_lin=False` (default). Patch-level (additive keyword-only kwarg).
-- **HAD R-package end-to-end parity test** vs `DIDHAD` v2.0.0 (`Credible-Answers/did_had`) on the **`design="continuous_at_zero"` (Design 1') surface**. New parity fixture `benchmarks/data/did_had_golden.json` generated by `benchmarks/R/generate_did_had_golden.R` covers 3 paper-derived synthetic DGPs (Uniform, Beta(2,2), Beta(0.5,1)) × 5 method combinations (overall, event-study, placebo, yatchew, trends_lin). The harness explicitly forces `HeterogeneousAdoptionDiD(design="continuous_at_zero")` because R `did_had` always evaluates the local-linear at `d=0` regardless of dose distribution; our default `design="auto"` may legitimately choose `continuous_near_d_lower` or `mass_point` on dose distributions with boundary density bounded away from zero (e.g., Beta(2,2)) and thereby diverge from R numerically — that divergence is methodologically defensible but out of scope for this parity test. Python parity test `tests/test_did_had_parity.py` asserts point estimate / SE / CI bounds at `atol=1e-8` and Yatchew T-stat at `atol=1e-10` after a documented `× G/(G-1)` finite-sample convention shift. Two intentional convention deviations from R, documented in `docs/methodology/REGISTRY.md`: (a) we report the bias-corrected point estimate (modern CCF 2018 convention; R's `Estimate` column reports the conventional estimate with the bias-corrected CI separately — our `att` matches R's CI midpoint); (b) Yatchew uses paper Appendix E's literal (1/G) variance-denominator convention while R uses base-R `var()`'s (1/(N-1)) sample-variance convention (parity is bit-exact after the `× G/(G-1)` shift). Yatchew on placebos with R's mean-independence null (`order=0`) is not yet exposed in our `yatchew_hr_test` (we currently only support the linearity null) and is skipped in the parity test; tracked as TODO follow-up.
+- **HAD R-package end-to-end parity test** vs `DIDHAD` v2.0.0 (`Credible-Answers/did_had`) on the **`design="continuous_at_zero"` (Design 1') surface**. New parity fixture `benchmarks/data/did_had_golden.json` generated by `benchmarks/R/generate_did_had_golden.R` covers 3 paper-derived synthetic DGPs (Uniform, Beta(2,2), Beta(0.5,1)) × 5 method combinations (overall, event-study, placebo, yatchew, trends_lin). The harness explicitly forces `HeterogeneousAdoptionDiD(design="continuous_at_zero")` because R `did_had` always evaluates the local-linear at `d=0` regardless of dose distribution; our default `design="auto"` may legitimately choose `continuous_near_d_lower` or `mass_point` on dose distributions with boundary density bounded away from zero (e.g., Beta(2,2)) and thereby diverge from R numerically — that divergence is methodologically defensible but out of scope for this parity test. Python parity test `tests/test_did_had_parity.py` asserts point estimate / SE / CI bounds at `atol=1e-8` and Yatchew T-stat at `atol=1e-10` after a documented `× G/(G-1)` finite-sample convention shift. Two intentional convention deviations from R, documented in `docs/methodology/REGISTRY.md`: (a) we report the bias-corrected point estimate (modern CCF 2018 convention; R's `Estimate` column reports the conventional estimate with the bias-corrected CI separately — our `att` matches R's CI midpoint); (b) Yatchew uses paper Appendix E's literal (1/G) variance-denominator convention while R uses base-R `var()`'s (1/(N-1)) sample-variance convention (parity is bit-exact after the `× G/(G-1)` shift). Yatchew on placebos with R's mean-independence null (`order=0`) was not exposed in `yatchew_hr_test` at the PR #392 cut and was skipped in the parity test; the follow-up `yatchew_hr_test(null="mean_independence")` entry above closes that gap (placebo rows now routed through `null="mean_independence"` and parity holds at the same `atol=1e-10`).
 - **Tutorial 20: HAD for National Brand Campaign with Regional Spend Intensity** (`docs/tutorials/20_had_brand_campaign.ipynb`) — end-to-end practitioner walkthrough for `HeterogeneousAdoptionDiD` on a 60-DMA panel where every market is treated at a different dose level and no never-treated unit exists; comparison comes from dose variation across markets, not from an untreated holdout. The DGP uses Uniform[\$5K, \$50K] regional add-on spend per DMA (every DMA participates, no DMA at exactly \$0), so `design="auto"` resolves to `continuous_near_d_lower` (Design 1) with target `WAS_d_lower` — interpreted as the average per-dollar marginal effect of regional spend above the lightest-touch DMA's spend (`d_lower` ≈ \$5K). Covers the headline `WAS_d_lower` fit on a 2-period collapse, the multi-week event study with per-week pointwise CIs and pre-launch placebos, and a stakeholder communication template that flags the Assumption 5/6 caveat (non-testable local-linearity at the boundary). Companion drift-test file `tests/test_t20_had_brand_campaign_drift.py` (13 tests pinning panel composition / sample median, design auto-detection / target / `d_lower`, overall `WAS_d_lower`, CI endpoints, dose mean, n_units, full event-study horizon presence, and per-horizon coverage). T20 wired into the existing `had.py` entry in `docs/doc-deps.yaml`; cross-link added from `docs/practitioner_decision_tree.rst` § "Universal Rollout (No Untreated Markets)" via a `.. tip::` block.
 
 ### Changed

TODO.md

@@ -103,7 +103,7 @@ Deferred items from PR reviews that were not addressed before merge.
 | `HeterogeneousAdoptionDiD` continuous paths: thread `cluster=` through `bias_corrected_local_linear` (Phase 1c's wrapper already supports cluster; Phase 2a ignores it with a `UserWarning` on the continuous path to keep scope tight). | `diff_diff/had.py`, `diff_diff/local_linear.py` | Phase 2a | Low |
 | `HeterogeneousAdoptionDiD` Eq 17 / Eq 18 linear-trend detrending: SHIPPED in PR #389 (Phase 4 R-parity, 2026-04). Exposed as `trends_lin: bool = False` keyword-only kwarg on `HeterogeneousAdoptionDiD.fit(aggregate="event_study")`, `joint_pretrends_test`, `joint_homogeneity_test`. Mirrors R `DIDHAD::did_had(..., trends_lin=TRUE)`. Pierce-Schott published-number parity (paper p=0.51 / p=0.40) deferred indefinitely (LBD-restricted analysis panel); replaced by end-to-end R-package parity at `tests/test_did_had_parity.py`. | `diff_diff/had_pretests.py::joint_pretrends_test`, `diff_diff/had.py` | Phase 4 (shipped) | Done |
 | `HeterogeneousAdoptionDiD` `trends_lin × survey_design` follow-up: per-group linear-trend slope under survey weighting (weighted slope estimator? per-PSU slope?) is not derived from the paper. PR #389 raises `NotImplementedError` on the combination across all 3 trends_lin surfaces. If user demand emerges, derive the weighted variant and lift the gate. | `diff_diff/had.py::HeterogeneousAdoptionDiD.fit`, `diff_diff/had_pretests.py::joint_pretrends_test`, `diff_diff/had_pretests.py::joint_homogeneity_test` | follow-up | Low |
-| `HeterogeneousAdoptionDiD` `yatchew_hr_test(null="mean_independence")` mode: R `YatchewTest::yatchew_test(order=0)` fits `Y ~ 1` (intercept-only baseline) and tests mean-independence of Y from D; R's `DIDHAD::did_had(yatchew=TRUE)` uses this on placebo rows ("non-parametric pre-trends test"). Our `yatchew_hr_test` always fits `Y ~ D` (linearity null) — no `null=` parameter exposed. Adding the mean-independence mode would (a) give practitioners a more conventional pre-trends test surface, and (b) close the PR #389 R-parity feature gap on the placebo-Yatchew rows (currently skipped in `tests/test_did_had_parity.py::TestYatchewParity` because the two tests are not the same statistic). | `diff_diff/had_pretests.py::yatchew_hr_test` | follow-up | Medium |
+| `HeterogeneousAdoptionDiD` `yatchew_hr_test(null="mean_independence")` mode: SHIPPED post-PR #392 (2026-04). Adds `null: Literal["linearity", "mean_independence"]` keyword-only kwarg mirroring R `YatchewTest::yatchew_test(order=0)`. Default `"linearity"` is bit-exact backcompat. `tests/test_did_had_parity.py::TestYatchewParity` now routes placebo rows through `null="mean_independence"` (R `order=0`) and effect rows through `null="linearity"` (R `order=1`); parity holds at `atol=1e-10` after the documented `× G/(G-1)` finite-sample convention shift. | `diff_diff/had_pretests.py::yatchew_hr_test` | follow-up (shipped) | Done |
 | `HeterogeneousAdoptionDiD` Stute family Stata-bridge parity: PR #389 R-parity covers the full HAD fit + Yatchew surfaces but skips Stute family (`stute_test`, `stute_joint_pretest`, `joint_pretrends_test`, `joint_homogeneity_test`) because no R `Stutetest` package exists publicly (chaisemartinPackages publishes only the Stata `stute_test` module; the paper cites a 2024c R Stutetest module that is not on GitHub or CRAN). Stata-bridge parity would add `benchmarks/stata/generate_stute_golden.do` + a Stata installation requirement. Low priority unless user demand emerges. | `benchmarks/stata/`, `tests/test_stute_test_parity.py` | follow-up | Low |
 | `HeterogeneousAdoptionDiD` Phase 3 Stute performance: Appendix D vectorized matrix form replaces the per-iteration OLS refit with a single precomputed `M = I - X(X'X)^{-1}X'` applied to `eps * eta`. Functionally identical, ~2x faster. Shipped literal-refit form in Phase 3 to match paper text and keep reviewer surface small. | `diff_diff/had_pretests.py::stute_test` | Phase 3 | Low |
 | `HeterogeneousAdoptionDiD` Phase 3 R-parity: Phase 3 ships coverage-rate validation on synthetic DGPs (not tight point parity against `chaisemartin::stute_test` / `yatchew_test`). Tight numerical parity requires aligning bootstrap seed semantics and `B` across numpy/R and is deferred. | `tests/test_had_pretests.py` | Phase 3 | Low |