Issues

Untrack the stray .claude/scheduled_tasks.lock

gregfaletto/fetwfePackage#542 · opened

on Oct 31, 2026

CRAN MKL check fails: exact comparisons between separately computed fits (deadline 2026-11-21)

gregfaletto/fetwfePackage#541 · opened

on Oct 31, 2026

2

Two passages recommend the bootstrap for any high-dimensional fit, not only fetwfe() fits

gregfaletto/fetwfePackage#538 · opened

on Oct 15, 2026

test-broom-methods.R's dispatch test can't see a lost S3 registration, though its comment says it can

gregfaletto/fetwfePackage#537 · opened

on Oct 12, 2026

Docs: say fetwfe() targets real-valued outcomes and that binary outcomes are out of scope

gregfaletto/fetwfePackage#536 · opened

on Oct 12, 2026

REML variance estimates depend on unit order at the optimizer's tolerance

gregfaletto/fetwfePackage#535 · opened

on Oct 12, 2026

1

Fixed-seed multiplier bootstrap depends on unit labels: relabeling moves the band

gregfaletto/fetwfePackage#534 · opened

on Oct 12, 2026

pkgdown site keeps orphaned pages: the deploy runs with clean: false

gregfaletto/fetwfePackage#533 · opened

on Oct 12, 2026

Coverage studies for the new inference functions (stage 2): debiasedATT(), simultaneousCIs(), cohortTimeATTs()

gregfaletto/fetwfePackage#532 · opened

on Oct 12, 2026

Docs-vs-behavior sweep: check every checkable claim in the help pages, vignettes and README against the code

gregfaletto/fetwfePackage#531 · opened

on Oct 12, 2026

Coverage simulation for the default fetwfe() fit, as a manually triggered CI job

gregfaletto/fetwfePackage#530 · opened

on Oct 12, 2026

Tests: invariance checks across all estimators

gregfaletto/fetwfePackage#529 · opened

on Oct 12, 2026

Tests: known-answer checks against etwfe, fixest and did, plus the no- and full-fusion limits

gregfaletto/fetwfePackage#528 · opened

on Oct 12, 2026

[Release] debiasedATT(): withheld from the next CRAN release; fix and validate it for re-export

gregfaletto/fetwfePackage#527 · opened

on Oct 12, 2026

1

Coverage test computes the true ATT from the simulator's independent counts, not the realized sample (and paper_coverage_writeup.tex's 0.933)

gregfaletto/fetwfePackage#526 · opened

on Oct 12, 2026

Bridge solver runs at grpreg's default eps = 1e-3: the no-fusion limit misses OLS/ETWFE

gregfaletto/fetwfePackage#525 · opened

on Oct 12, 2026

fetwfe registers S3 methods for class "etwfe", breaking fits from the etwfe package

gregfaletto/fetwfePackage#523 · opened

on Oct 12, 2026

1

Convert the remaining British spellings to American, and give NEWS.md's development header its version

gregfaletto/fetwfePackage#522 · opened

on Oct 10, 2026

1

Stale out-list field counts in result_assembly.R and test-build-selected-out-result-80.R

gregfaletto/fetwfePackage#521 · opened

on Oct 10, 2026

Fixed-p debiasedATT()'s var_reg moves with the route; the OLS residual removes that but under-covers at moderate p/N without a small-sample correction

gregfaletto/fetwfePackage#520 · opened

on Oct 10, 2026

debiasedATT()'s cohort-weight SE channel (V2) still plugs in the fused fit's cohort estimates, so the fixed-p SE moves with the route

gregfaletto/fetwfePackage#518 · opened

on Oct 10, 2026

test-plot-methods-29.R's 'ggplot2 missing' test raises its own error and never calls package code

gregfaletto/fetwfePackage#517 · opened

on Oct 10, 2026

2

?cohortStudy and ?tidy.fetwfe say a bridge-zeroed cohort's SE is NA, but measured fits store 0

gregfaletto/fetwfePackage#516 · opened

on Oct 10, 2026

1

eventStudy() reports estimate = 0 at an event time no cohort reaches, and plot() draws it as a zero effect

gregfaletto/fetwfePackage#515 · opened

on Oct 10, 2026

1

debiasedATT(): lift the whitened q >= 1 fixed-p refusal, which #507 left without a rationale, and add a method-entry precondition

gregfaletto/fetwfePackage#513 · opened

on Oct 10, 2026

2

Navigation Menu

Repository navigation

Issues

Untrack the stray .claude/scheduled_tasks.lock

CRAN MKL check fails: exact comparisons between separately computed fits (deadline 2026-11-21)

Two passages recommend the bootstrap for any high-dimensional fit, not only fetwfe() fits

test-broom-methods.R's dispatch test can't see a lost S3 registration, though its comment says it can

Docs: say fetwfe() targets real-valued outcomes and that binary outcomes are out of scope

REML variance estimates depend on unit order at the optimizer's tolerance

Fixed-seed multiplier bootstrap depends on unit labels: relabeling moves the band

pkgdown site keeps orphaned pages: the deploy runs with clean: false

Coverage studies for the new inference functions (stage 2): debiasedATT(), simultaneousCIs(), cohortTimeATTs()

Docs-vs-behavior sweep: check every checkable claim in the help pages, vignettes and README against the code

Coverage simulation for the default fetwfe() fit, as a manually triggered CI job

Tests: invariance checks across all estimators

Tests: known-answer checks against etwfe, fixest and did, plus the no- and full-fusion limits

[Release] debiasedATT(): withheld from the next CRAN release; fix and validate it for re-export

Coverage test computes the true ATT from the simulator's independent counts, not the realized sample (and paper_coverage_writeup.tex's 0.933)

Bridge solver runs at grpreg's default eps = 1e-3: the no-fusion limit misses OLS/ETWFE

fetwfe registers S3 methods for class "etwfe", breaking fits from the etwfe package

Convert the remaining British spellings to American, and give NEWS.md's development header its version

Stale out-list field counts in result_assembly.R and test-build-selected-out-result-80.R

Fixed-p debiasedATT()'s var_reg moves with the route; the OLS residual removes that but under-covers at moderate p/N without a small-sample correction

debiasedATT()'s cohort-weight SE channel (V2) still plugs in the fused fit's cohort estimates, so the fixed-p SE moves with the route

test-plot-methods-29.R's 'ggplot2 missing' test raises its own error and never calls package code

?cohortStudy and ?tidy.fetwfe say a bridge-zeroed cohort's SE is NA, but measured fits store 0

eventStudy() reports estimate = 0 at an event time no cohort reaches, and plot() draws it as a zero effect

debiasedATT(): lift the whitened q >= 1 fixed-p refusal, which #507 left without a rationale, and add a method-entry precondition