diff --git a/CHANGELOG.md b/CHANGELOG.md index 323c19c..ba2f660 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,6 +1,6 @@ # Changelog -## 0.3.0 (unreleased): LP-DiD, negative horizons, lead paths +## 0.2.1 (2026-11-18): LP-DiD, negative horizons, lead paths ### LP-DiD (Dube, Girardi, Jordà & Taylor 2025) - `estimator = LPDiD(…)` with the formula `ldiff(y) ~ firstdiff(D) + fe(time)`: the long diff --git a/CITATION.bib b/CITATION.bib index ad8fdd4..e80e418 100644 --- a/CITATION.bib +++ b/CITATION.bib @@ -3,7 +3,7 @@ @software{PanelLocalProjections_jl title = {{PanelLocalProjections.jl}: Panel local projections in {Julia}}, url = {https://github.com/andrerecio/PanelLocalProjections.jl}, year = {2026}, - version = {0.3.0} + version = {0.2.1} } @article{MeiShengShi2026, diff --git a/CLAUDE.md b/CLAUDE.md index a4a1f51..7324f69 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -11,7 +11,7 @@ PanelLocalProjections.jl estimates panel local projections: for each horizon h i - **Dube–Girardi–Jordà–Taylor (2025, *JAE*):** **LP-DiD**, difference-in-differences by local projections on clean-control samples. Their Stata command is `lpdid` (GPL-3). - **Almuzara–Sancibrián (2026):** **t-LAHR** inference. It is time-clustered, lag-augmented and heteroskedasticity-robust, with an HC2 + Imbens–Kolesár df refinement. `starting_idea.md` calls it a "HAR estimator", but it is **not** HAR: Driscoll–Kraay, EWC and unit/two-way clustering are its *comparators*. -The phased roadmap, with exit criteria per phase, is in `TODO.md`. Panel LP-IV is v0.2.0 (Phase 6: 6a IV core 19bf450, 6b lag-augmented IV 91564d3, 6c first-stage diagnostics, a public-data example and the release; design in `docs/notes/design/lpiv_design.md`). LP-DiD, negative horizons, lead paths, weights and pooled fits are v0.3.0 (Phase 7). +The phased roadmap, with exit criteria per phase, is in `TODO.md`. Panel LP-IV is v0.2.0 (Phase 6: 6a IV core 19bf450, 6b lag-augmented IV 91564d3, 6c first-stage diagnostics, a public-data example and the release; design in `docs/notes/design/lpiv_design.md`). LP-DiD, negative horizons, lead paths, weights and pooled fits are v0.2.1 (Phase 7). ## Hard rules @@ -69,7 +69,7 @@ Rscript crosscheck/R/.R - **Kaimon.** When the `mcp__kaimon__*` tools are available, use the live REPL (session for this repo via `start_session(project_path = …)`, then `ex`) instead of spawning `julia` for exploration. Revise picks up `src/` edits automatically, but a struct change in `src/types.jl` needs `manage_repl(command = "restart")`. Load test-only packages and the fixtures with `pushfirst!(LOAD_PATH, joinpath(pkgdir(PanelLocalProjections), "dev"))` and `include("test/support/PLPDGPs.jl")` (`read_csv`). `run_tests` runs the whole suite, because TestItemRunner ignores its pattern argument. - **Scale benchmark:** `julia --project=. -e 'push!(LOAD_PATH, "dev"); include("dev/benchmark_scale.jl")'` (about 70 s; `PLP_BENCH=fe`, `tlahr` or `capped` runs one part, for its own max RSS). Reference for 1.1M rows and 21 horizons: fixed effects 13.9 s, 14 sweeps per horizon, 1.3 GB max RSS; `LagAugmented(ICLags())` 19.6 s and `TLAHR()` 0.8 s, 1.8 GB; `LagAugmented(CappedLags())` 24.5 s, 2.2 GB (Phase 5). -## Architecture (v0.1: Phases 1–5; v0.2: Phase 6, panel LP-IV; v0.3: Phase 7, LP-DiD; see `TODO.md`) +## Architecture (v0.1: Phases 1–5; v0.2: Phase 6, panel LP-IV; v0.2.1: Phase 7, LP-DiD; see `TODO.md`) One module and no submodules. Every struct lives in `src/types.jl`. Include order: `utils` → `types` → `panelindex` → `formula` → `design` → `absorb` → `ols` → `scores` → `lagselect` → `estimators` → `panellp` → `vcov` → `summary` → `lpdid` → `plotting`; Makie lives in `ext/`. diff --git a/Project.toml b/Project.toml index a70fac0..9ec1507 100644 --- a/Project.toml +++ b/Project.toml @@ -1,7 +1,7 @@ name = "PanelLocalProjections" uuid = "4bc008c6-1160-4c54-8f8b-e57ec2b93f89" authors = ["Andrea Recine "] -version = "0.3.0" +version = "0.2.1" [deps] Dates = "ade2ca70-3891-5945-98fb-dc099432e06a" diff --git a/README.md b/README.md index 2bf7378..12e6179 100644 --- a/README.md +++ b/README.md @@ -16,7 +16,7 @@ procedures of the recent panel-LP literature in one place: | LP-DiD: difference-in-differences by local projections on clean-control samples, for absorbing, non-absorbing and one-off treatments, with reweighting, regression adjustment, pre-mean differencing, pooled estimates and a pre-trend test | Dube, Girardi, Jordà & Taylor (2025, *JAE*) | | Pre-trends (negative horizons) and the response holding the later path of the treatment fixed, for a continuous, instrumented treatment | Ilzetzki (2024, *AER*) | -Version 0.3.0, for Julia 1.10 (LTS) and later. What comes after it (joint bands across +Version 0.2.1, for Julia 1.10 (LTS) and later. What comes after it (joint bands across horizons, weak-IV-robust inference) is listed in [`TODO.md`](TODO.md); the changes of each version are in [`CHANGELOG.md`](CHANGELOG.md). The [documentation](#documentation) covers the methods, the conventions of each reference @@ -30,7 +30,7 @@ The package is not in the General registry. Install it from GitHub: ```julia using Pkg Pkg.add(url = "https://github.com/andrerecio/PanelLocalProjections.jl") # main -Pkg.add(url = "https://github.com/andrerecio/PanelLocalProjections.jl", rev = "v0.2.0") # or a tagged release +Pkg.add(url = "https://github.com/andrerecio/PanelLocalProjections.jl", rev = "v0.2.1") # or a tagged release ``` It depends on StatsModels, Tables, StatsAPI, StatsFuns, RecipesBase and Julia's standard diff --git a/TODO.md b/TODO.md index f08365a..41bb192 100644 --- a/TODO.md +++ b/TODO.md @@ -208,7 +208,7 @@ structural residual. Every existing variance then applies unchanged, and `TLAHR( Non-IV and IV fits are bit-identical to 6b. - [x] Phase 6c gate; commit -## Phase 7: LP-DiD, negative horizons, lead paths (v0.3) +## Phase 7: LP-DiD, negative horizons, lead paths (v0.2.1) References: Dube, Girardi, Jordà & Taylor (2025, *JAE*) and the Stata command `lpdid` 1.1.0 (GPL-3: run as a black box, never ported); the replication package of Ilzetzki (2024, AER; @@ -272,7 +272,7 @@ BSD-3 code, CC BY 4.0 data, local only). - [x] Phase 7 gate (2026-11-18): `Pkg.test()` on Julia 1.13.3 and 1.10.13 (45,623 tests, the local MSS and Ilzetzki items included), format, fixtures, docs build, the p7 Stata fixtures regenerated byte-identically, ado manifest -- [ ] Commit (awaits the user's go-ahead) +- [x] Commit (93f0229; merged to `main` by PR #2, 3d31caf); released as v0.2.1 ## After v0.1 - [ ] Adversarial review of the Phase 4 review fixes and of the Phase 5 memory changes (deferred; started twice, stopped for time). Scope: diff --git a/docs/src/index.md b/docs/src/index.md index 14ad0de..c0fd616 100644 --- a/docs/src/index.md +++ b/docs/src/index.md @@ -24,7 +24,7 @@ Panel local projections (Jordà 2005): for each horizon ``h``, a panel regressio ([replication](replication/ilzetzki2024.md)). !!! note "Status" - Version 0.3.0. The roadmap is in `TODO.md` at the repository root. + Version 0.2.1. The roadmap is in `TODO.md` at the repository root. Fixed-effect absorption, the OLS solver and all variance estimators are implemented in the package. Their design takes inspiration from diff --git a/docs/src/replication/ilzetzki2024.md b/docs/src/replication/ilzetzki2024.md index 7d9d7ea..3629e30 100644 --- a/docs/src/replication/ilzetzki2024.md +++ b/docs/src/replication/ilzetzki2024.md @@ -4,7 +4,7 @@ Ilzetzki (2024, *American Economic Review*) estimates the response of aircraft p productivity to demand in World War II by panel LP-IV: 143 production lines, monthly, 1940–1945, with the output of a line instrumented by the output of the other lines of its broad aircraft type. Two features of his regressions are general options of `panellp` -since v0.3 ([Lead paths and negative horizons](../methods/leads.md)): +since v0.2.1 ([Lead paths and negative horizons](../methods/leads.md)): - **pre-trends**: the loop runs from a negative horizon (``h = -6`` in his Figure A.4, ``-1`` elsewhere), with the outcome ``y_{i,t+h} - y_{i,t-1}`` at every ``h``;