Address CI review P3: broaden B-spline derivative warning scope

CI reviewer correctly noted that dPsi feeds both ACRT point estimates
(continuous_did.py:1026-1046, bootstrap path at 1524-1561) and the
analytical/bootstrap inference, not just inference alone. Updated the
warning text and the REGISTRY note to say both outputs are biased when
derivative construction fails.

No behavior change; text-only accuracy fix.

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

Author: igerber

diff_diff/continuous_did_bspline.py

@@ -172,9 +172,10 @@ def bspline_derivative_design_matrix(x, knots, degree, include_intercept=True):
                 f"(indices {failed_basis_indices}); their derivative columns "
                 f"are zero. This typically indicates a malformed knot vector "
                 f"(too few knots for the chosen degree, non-monotonic, or "
-                f"repeated interior knots). ContinuousDiD inference may be "
-                f"biased; consider increasing the number of distinct doses "
-                f"or reducing the B-spline degree.",
+                f"repeated interior knots). Both ACRT point estimates and "
+                f"analytical/bootstrap inference depend on this derivative "
+                f"matrix, so both may be biased. Consider increasing the "
+                f"number of distinct doses or reducing the B-spline degree.",
                 UserWarning,
                 stacklevel=2,
             )

docs/methodology/REGISTRY.md

@@ -723,7 +723,7 @@ See `docs/methodology/continuous-did.md` Section 4 for full details.
   not-yet-treated controls. When `anticipation=0` (default), behavior is
   unchanged.
 - **Boundary knots**: Knots are built once from all treated doses (global, not per-cell) to ensure a common basis across (g,t) cells for aggregation. Evaluation grid is clamped to training-dose boundary knots (`range(dose)`). R's `contdid` v0.1.0 has an inconsistency where `splines2::bSpline(dvals)` uses `range(dvals)` instead of `range(dose)`, which can produce extrapolation artifacts at dose grid extremes. Our approach avoids extrapolation and is methodologically sound.
-- **Note:** `bspline_derivative_design_matrix` previously swallowed `ValueError` from `scipy.interpolate.BSpline` in the per-basis derivative loop, leaving affected columns of the derivative design matrix as zero with no user-facing signal. It now aggregates the failed basis indices and emits ONE `UserWarning` naming them so downstream ContinuousDiD inference doesn't silently use a biased `dPsi`. The all-identical-knot degenerate case (single dose value) remains silently handled — derivatives there are mathematically zero. Axis-C finding #12 in the Phase 2 silent-failures audit.
+- **Note:** `bspline_derivative_design_matrix` previously swallowed `ValueError` from `scipy.interpolate.BSpline` in the per-basis derivative loop, leaving affected columns of the derivative design matrix as zero with no user-facing signal. It now aggregates the failed basis indices and emits ONE `UserWarning` naming them. Both ACRT point estimates and analytical/bootstrap inference read the same `dPsi` matrix (see `continuous_did.py:1026-1046` and the bootstrap ACRT path at `continuous_did.py:1524-1561`), so both are biased on a partial derivative-construction failure — the warning wording makes that explicit. The all-identical-knot degenerate case (single dose value) remains silently handled — derivatives there are mathematically zero. Axis-C finding #12 in the Phase 2 silent-failures audit.
 
 ### Implementation Checklist