Fix HonestDiD identified set and inference (methodology review)#248

Merged

igerber merged 16 commits into

mainfrom

honest-did-validation

Apr 1, 2026

Commits on Apr 1, 2026

Fix HonestDiD identified set computation and inference (F1-F6)

igerber
and
claude
committed
Optimize FLCI performance: Newton cv_alpha, centrosymmetric LP, M=0 short-circuit

igerber
and
claude
committed
Fix REGISTRY.md HonestDiD equations: second diffs, first diffs, LP constraints

igerber
and
claude
committed
Disable ARP grid search for RM CI (use naive FLCI while calibrating)

igerber
and
claude
committed
Add methodology verification tests for HonestDiD (17 tests)

igerber
and
claude
committed
Update METHODOLOGY_REVIEW.md: HonestDiD review complete

igerber
and
claude
committed
Address AI review findings: fix FLCI bias, LP infeasibility, survey df, labels

igerber
and
claude
committed
Address re-review: smoothness df gating, M=0 SE, infeasibility propagation

igerber
and
claude
committed
Rewrite optimal FLCI with first-difference reparameterization

igerber
and
claude
committed
Fix remaining stale RM strings and update review status

igerber
and
claude
committed
Clarify METHODOLOGY_REVIEW.md: DeltaSD uses proper optimal FLCI

igerber
and
claude
committed
Fix FLCI slope count (T not T-1) and constraint (sum_j j*l_j not 1)

igerber
and
claude
committed
Document combined-mode FLCI inconsistency, fix stale comments, optimize RM LP

igerber
and
claude
committed
Fix NaN CI misclassification as significant (P0)

igerber
and
claude
committed
Fix REGISTRY.md: M=0 linear extrapolation, survey df folded nct

igerber
and
claude
committed
Relax M=0 timing assertion for CI runners (0.1s -> 0.5s)

igerber
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix HonestDiD identified set and inference (methodology review)#248

Fix HonestDiD identified set and inference (methodology review)#248
igerber merged 16 commits into
mainfrom
honest-did-validation

Commits on Apr 1, 2026

Fix HonestDiD identified set computation and inference (F1-F6)

Optimize FLCI performance: Newton cv_alpha, centrosymmetric LP, M=0 short-circuit

Fix REGISTRY.md HonestDiD equations: second diffs, first diffs, LP constraints

Disable ARP grid search for RM CI (use naive FLCI while calibrating)

Add methodology verification tests for HonestDiD (17 tests)

Update METHODOLOGY_REVIEW.md: HonestDiD review complete

Address AI review findings: fix FLCI bias, LP infeasibility, survey df, labels

Address re-review: smoothness df gating, M=0 SE, infeasibility propagation

Rewrite optimal FLCI with first-difference reparameterization

Fix remaining stale RM strings and update review status

Clarify METHODOLOGY_REVIEW.md: DeltaSD uses proper optimal FLCI

Fix FLCI slope count (T not T-1) and constraint (sum_j j*l_j not 1)

Document combined-mode FLCI inconsistency, fix stale comments, optimize RM LP

Fix NaN CI misclassification as significant (P0)

Fix REGISTRY.md: M=0 linear extrapolation, survey df folded nct

Relax M=0 timing assertion for CI runners (0.1s -> 0.5s)