feat: add covariate support to ChangesInChanges/QDiD (qte xformla parity)

.github/workflows/ai_pr_review.yml

@@ -340,9 +340,13 @@ jobs:
             echo "Unified diff (context=5):"
             # Exclude large generated/data files from the full diff to stay
             # within the model's input limit. The --name-status above still
-            # lists them. Narrowed to real-data assets and notebook outputs.
+            # lists them. Narrowed to real-data assets, notebook outputs, and
+            # the qte golden fixture (generated single-line JSON, ~700KB - a
+            # regeneration blew the input limit at PR #688; the reviewable
+            # artifact is benchmarks/R/generate_qte_golden.R).
             git --no-pager diff --unified=5 "$BASE_SHA" "$HEAD_SHA" \
               -- . ':!benchmarks/data/real/*.json' ':!benchmarks/data/real/*.csv' \
+                   ':!benchmarks/data/qte_golden.json' \
                    ':!docs/tutorials/*.ipynb'
           } >> "$PROMPT"
 

CHANGELOG.md

@@ -83,20 +83,36 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   DiD for the canonical 2x2 design with continuous outcomes: `ChangesInChanges`
   (alias `CiC`) recovers the treated group's full counterfactual outcome
   distribution and quantile treatment effects via the CDF transformation
-  `F_10(F_00^{-1}(F_01(y)))`, invariant to monotone outcome rescaling; `QDiD`
+  `F_10(F_00^{-1}(F_01(y)))`, invariant to monotone outcome rescaling
+  (unconditional fits; the covariate QR branch is not); `QDiD`
   is the paper's additive quantile-DiD comparison estimator. Point estimation
   matches R `qte` 1.3.1 exactly (CiC via an exact port of R's type-1 quantile
   arithmetic incl. its fuzz handling - bit-exact golden parity; QDiD via qte's
   type-7 form, documented in REGISTRY.md as a finite-sample deviation from the
   paper's transformation, population-equivalent). Bootstrap-only inference
   (panel unit-block or pooled repeated-cross-section resampling per qte's
   schemes; seedable, `n_bootstrap=200` default; replicate-SD SEs, normal
-  intervals, sup-t uniform bands at qte's fixed 95% level). Diagnostics:
-  interior-range guard (eq. 17) with NaN inference outside, support-violation /
-  heavy-ties / QDiD-non-monotonicity warnings. Deferred and documented:
-  covariates, discrete-outcome bounds, analytical SEs, staggered designs.
-  Golden fixtures: `benchmarks/R/generate_qte_golden.R` (qte pinned 1.3.1) ->
-  `benchmarks/data/qte_golden.json`.
+  intervals, sup-t uniform bands at qte's fixed 95% level). Covariates are
+  supported on both estimators (`covariates=[...]` or trailing formula terms),
+  porting qte's `xformla` branch exactly: per-cell linear quantile regressions
+  (Koenker-Bassett LP via scipy HiGHS) on qte's fixed internal 0.01-0.99 tau
+  grid, quantreg `predict.rqs` Fhat/Qhat step-function conventions verbatim,
+  per-observation conditional-rank imputation, QR refits inside every
+  bootstrap replicate, and a conditional-envelope support diagnostic
+  (Melly-Santangelo 2015, Assumption 4); the unconditional interior-range
+  guard applies to covariate-free fits only. Diagnostics: interior-range
+  guard (eq. 17) with NaN inference outside, support-violation / heavy-ties /
+  QDiD-non-monotonicity warnings (the latter two scoped as documented in
+  REGISTRY.md under covariates). Deferred and documented: the full
+  Melly-Santangelo covariate estimator, discrete-outcome bounds, analytical
+  SEs, staggered designs. Golden fixtures:
+  `benchmarks/R/generate_qte_golden.R` (qte pinned 1.3.1, quantreg 6.1) ->
+  `benchmarks/data/qte_golden.json`, with covariate scenarios, atol=0
+  convention micro-fixtures (`qr_cases`), and deterministic seeded SE blocks
+  (`cores=1`; qte's default forked bootstrap is not reproducible even seeded).
+  Cross-language covariate parity is layered - conventions bit-exact given R
+  inputs, LP solver proven optimal per tau, end-to-end results gated at
+  documented exact-tie selection bounds (see the REGISTRY Deviation note).
 
 ### Changed
 - `diff_diff/guides/llms-autonomous.txt` no longer lists regression discontinuity as

ROADMAP.md

@@ -84,12 +84,6 @@ Extends DiD to duration / survival outcomes where standard methods fail (hazard
 
 **Reference**: Deaner & Ku (2025), *AEA Conference Paper*.
 
-### Quantile / Distributional DiD
-
-Recover the full counterfactual distribution and quantile treatment effects (QTT), not just mean ATT. Changes-in-Changes (CiC) identification strategy.
-
-**Reference**: Athey & Imbens (2006), *Econometrica*. (Ciaccio 2024 extension listed under Under Consideration.)
-
 ### CATT Meta-Learner for Heterogeneous Effects
 
 ML-powered conditional ATT, using a doubly robust meta-learner to discover which units benefit most from treatment.

TODO.md

@@ -41,6 +41,7 @@ generic sparse-FE, QR+SVD rank-detection redundancy, `check_finite` bypass — m
 | Issue | Location | Origin | Effort | Priority |
 |-------|----------|--------|--------|----------|
 | `EfficientDiD` conditional path: the largest remaining O(n) stage is the sieve/nuisance construction outside the tiled pass (~9s at 10k). (The `_ridge_solve_weights` Python-prep shave landed 2026-07-07 — the `omega_stack[rest]` fancy-index copy and tail scatter are skipped when no row is zero-masked, byte-identical outputs; the `zero_mask` abs scan itself remains, needed for correctness.) | `efficient_did_covariates.py` | CS-scaling | Mid | Low |
+| `_rq_fit` LP assembly is dense (`A_eq = [X, I, -I]` with dense identity blocks, rebuilt per cell fit): a `scipy.sparse` construction would cut memory and likely HiGHS time for large cells / bootstrap-heavy covariate CiC/QDiD fits. CAVEAT before doing it: a different matrix representation can change HiGHS's vertex selection at degenerate/tied QR optima - end-to-end covariate goldens are tie-selection-gated (fine), but the `qr_cases` tight coefficient matches may shift to the equal-loss branch; re-run the parity suite and re-calibrate if needed. | `diff_diff/changes_in_changes.py::_rq_fit` | covariates PR | Quick | Low |
 | Evaluate flipping `DIFF_DIFF_SOLVE_OLS_FASTPATH` default-ON after an opt-in soak (the 2026-07 certified normal-equations Cholesky fast path, both backends). A flip needs: golden/parity-suite recapture at the tol-bounded posture (fitted ~1e-8 abs / SE ~1e-6 rel — the default today is byte-pinned in several benchmark conventions), certification-rate telemetry across real workloads (any decline is silent-correct but forfeits the speedup), and the staged default-flip protocol used for `df_convention` (v4-class change). | `diff_diff/linalg.py::_resolve_solve_ols_fastpath`, `rust/src/linalg.rs::solve_ols_chol` | CS-scaling | Mid | Low |
 
 ### Testing / docs
@@ -112,7 +113,7 @@ Doable in principle, but no current caller and/or explicitly out of paper scope.
 
 | Issue | Location | PR | Priority |
 |-------|----------|----|----------|
-| ChangesInChanges covariates (Melly-Santangelo 2015 QR pipeline: per-cell quantile-regression conditional CDFs -> conditional CiC -> integrate over treated-group covariates). No R parity target exists (the MS Stata code is the only implementation; distinct from Kranker's `cic`); would need simulation-based validation. Reviewed: `docs/methodology/papers/melly-santangelo-2015-review.md`. | `diff_diff/changes_in_changes.py` | #682 | Low |
+| ChangesInChanges FULL Melly-Santangelo covariate estimator (monotonized integrated-indicator conditional CDFs, treated-post `F_{X|11}` integration, exchangeable bootstrap with variance-weighted KS bands, tail trimming, pre-period specification test). The qte-`xformla` simplified form of the MS pipeline SHIPPED in the covariates PR (`covariates=` on both estimators, parity-tested vs qte 1.3.1); this row's earlier "No R parity target exists" claim was WRONG and is corrected in the MS review doc. The full estimator has no reference implementation (the MS Stata code is the only one; distinct from Kranker's `cic`) and would need simulation-based validation. Reviewed: `docs/methodology/papers/melly-santangelo-2015-review.md`. | `diff_diff/changes_in_changes.py` | #682 | Low |
 | ChangesInChanges discrete-outcome bounds + DCIC point identification (Athey-Imbens Sections 4/5.2 incl. Imbens-Manski intervals; Kranker's Stata `cic` is the reference). The shipped ties warning marks the boundary of the continuous scope. | `diff_diff/changes_in_changes.py` | #682 | Low |
 | ChangesInChanges analytical SEs (Athey-Imbens Theorems 5.1-5.3 influence functions, panel 5.5-5.7, Appendix B covariances; needs the footnote-31 boundary density estimator - note the review's suspected half-range/midpoint typo). Bootstrap is the shipped inference. | `diff_diff/changes_in_changes.py` | #682 | Low |
 | Staggered/multi-period distributional DiD (Athey-Imbens Section 6 / Ciaccio arXiv:2408.01208v2; `ecic` is the staggered event-study CiC lineage - a distinct method from Ciaccio's copula approach, do not conflate). Reviewed: `docs/methodology/papers/ciaccio-2024-review.md`; ROADMAP row is reviewed-deferred pending demand. | `diff_diff/changes_in_changes.py` | #682 | Low |