Add Phase 2 multi-horizon event study DID_l for ChaisemartinDHaultfoeuille

Implements ROADMAP items 2a-2h: multi-horizon DID_l via per-group
DID_{g,l} building block (Eq 3 of dynamic paper), per-horizon
analytical SE, dynamic placebos DID^{pl}_l, normalized DID^n_l,
cost-benefit aggregate delta, sup-t simultaneous confidence bands,
plot_event_study() integration, and R DIDmultiplegtDYN parity tests
at multiple horizons. L_max parameter controls multi-horizon mode;
L_max=None preserves exact Phase 1 behavior.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: igerber

CHANGELOG.md

@@ -21,6 +21,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **`twowayfeweights()`** — standalone helper function for the TWFE decomposition diagnostic (Theorem 1 of de Chaisemartin & D'Haultfœuille 2020), available without instantiating the full estimator. Returns a `TWFEWeightsResult` with per-cell weights, fraction negative, `sigma_fe`, and `beta_fe`.
 - **`generate_reversible_did_data()`** — new generator in `diff_diff.prep` producing reversible-treatment panel data for testing and tutorials. Patterns: `single_switch` (default, A5-safe), `joiners_only`, `leavers_only`, `mixed_single_switch`, `random`, `cycles`, `marketing`. Returns columns `group`, `period`, `treatment`, `outcome`, `true_effect`, `d_lag`, `switcher_type`.
 - **REGISTRY.md `## ChaisemartinDHaultfoeuille` section** — single canonical source for dCDH methodology, equations, edge cases, and all documented deviations from the R `DIDmultiplegtDYN` reference implementation. Cites the AER 2020 paper and the dynamic companion paper (NBER WP 29873) by reference; primary papers are upstream sources, not in-repo files.
+- **Phase 2: Multi-horizon event study for `ChaisemartinDHaultfoeuille`** — adds `L_max` parameter to `fit()` for computing `DID_l` at horizons `l = 1, ..., L_max` using the per-group building block from Equation 3 of the dynamic companion paper. Ships:
+  - Per-horizon point estimates and cohort-recentered analytical SE
+  - Dynamic placebos `DID^{pl}_l` with dual eligibility condition (Web Appendix Section 1.1)
+  - Normalized estimator `DID^n_l = DID_l / delta^D_l` (Section 3.2)
+  - Cost-benefit aggregate `delta` (Section 3.3, Lemma 4) — becomes `overall_att` when `L_max > 1`
+  - Sup-t simultaneous confidence bands via multiplier bootstrap
+  - `plot_event_study()` integration with `<50%` switcher warning for far horizons
+  - `to_dataframe(level="event_study")` and `to_dataframe(level="normalized")` output
+  - Per-horizon bootstrap with bootstrap SE/CI/p-value propagation to event_study_effects
+  - `L_max=None` (default) preserves exact Phase 1 behavior
 
 ## [3.0.1] - 2026-04-07
 

README.md

@@ -1213,6 +1213,32 @@ ChaisemartinDHaultfoeuille(
 | `n_groups_dropped_crossers`, `n_groups_dropped_singleton_baseline` | Filter counts (multi-switch groups dropped before estimation; singleton-baseline groups excluded from variance) |
 | `n_groups_dropped_never_switching` | Backwards-compatibility metadata. Never-switching groups participate in the variance via stable-control roles; this field is no longer a filter count. |
 
+**Multi-horizon event study** (Phase 2 - pass `L_max` to `fit()`):
+
+```python
+results = est.fit(data, outcome="outcome", group="group",
+                  time="period", treatment="treatment", L_max=5)
+
+# Per-horizon effects with analytical SE
+for horizon in sorted(results.event_study_effects):
+    e = results.event_study_effects[horizon]
+    print(f"  l={horizon}: DID_l={e['effect']:.3f} (SE={e['se']:.3f})")
+
+# Cost-benefit delta (becomes overall_att when L_max > 1)
+print(f"Cost-benefit delta: {results.cost_benefit_delta['delta']:.3f}")
+
+# Normalized effects: DID^n_l = DID_l / l (for binary treatment)
+for horizon in sorted(results.normalized_effects):
+    print(f"  DID^n_{horizon} = {results.normalized_effects[horizon]['effect']:.3f}")
+
+# Event study DataFrame (includes placebos as negative horizons)
+df = results.to_dataframe("event_study")
+
+# Plot (integrates with plot_event_study)
+from diff_diff import plot_event_study
+plot_event_study(results)
+```
+
 **Standalone TWFE decomposition diagnostic** (without fitting the full estimator):
 
 ```python
@@ -1232,7 +1258,7 @@ print(f"sigma_fe (sign-flipping threshold): {diagnostic.sigma_fe:.3f}")
 
 > **Note:** Phase 1 requires panels with a **balanced baseline** (every group observed at the first global period) and **no interior period gaps**. Late-entry groups (missing the baseline) raise `ValueError`; interior-gap groups are dropped with a warning; terminally-missing groups (early exit / right-censoring) are retained and contribute from their observed periods only. This is a documented deviation from R `DIDmultiplegtDYN`, which supports unbalanced panels — see [`docs/methodology/REGISTRY.md`](docs/methodology/REGISTRY.md) for the rationale, the defensive guards that make terminal missingness safe, and workarounds for unbalanced inputs.
 
-> **Note:** Survey design (`survey_design`), event-study aggregation (`aggregate`), covariate adjustment (`controls`), and HonestDiD integration (`honest_did`) are not yet supported. They raise `NotImplementedError` with phase pointers — see [`ROADMAP.md`](ROADMAP.md) for the full multi-phase rollout.
+> **Note:** Survey design (`survey_design`), covariate adjustment (`controls`), group-specific linear trends (`trends_linear`), and HonestDiD integration (`honest_did`) are not yet supported. They raise `NotImplementedError` with phase pointers - see [`ROADMAP.md`](ROADMAP.md) for the Phase 3 rollout.
 
 ### Triple Difference (DDD)
 

ROADMAP.md

@@ -148,18 +148,18 @@ The dynamic companion paper subsumes the AER 2020 paper: `DID_1 = DID_M`. The si
 
 ### Phase 2: Dynamic event study (multiple horizons)
 
-*Goal: Add `aggregate="event_study"` mode to the same class. Loops the Phase 1 machinery over horizons `l = 1, ..., L`. No API breakage from Phase 1. No new tutorial — the comprehensive tutorial waits for Phase 3.*
+*Goal: Add multi-horizon event study to the same class via the `L_max` parameter. Loops the Phase 1 machinery over horizons `l = 1, ..., L`. No API breakage from Phase 1. No new tutorial - the comprehensive tutorial waits for Phase 3.*
 
 | Item | Priority | Status |
 |------|----------|--------|
-| **2a.** Multi-horizon `DID_l` via the cohort framework, with horizon parameter `L_max` | HIGH | Not started |
-| **2b.** Multi-horizon analytical SE (same plug-in formula looped over horizons) | HIGH | Not started |
-| **2c.** Dynamic placebos `DID^{pl}_l` for pre-trends testing (Web Appendix Section 1.1 of dynamic paper) | HIGH | Not started |
-| **2d.** Normalized estimator `DID^n_l` (Section 3.2 of dynamic paper) | MEDIUM | Not started |
-| **2e.** Cost-benefit aggregate `delta` (Section 3.3 of dynamic paper, Lemma 4) | MEDIUM | Not started |
-| **2f.** Simultaneous (sup-t) confidence bands for event study plots | MEDIUM | Not started |
-| **2g.** `plot_event_study()` integration; `< 50%`-of-switchers warning for far horizons | MEDIUM | Not started |
-| **2h.** Parity tests vs `did_multiplegt_dyn` for multi-horizon designs | HIGH | Not started |
+| **2a.** Multi-horizon `DID_l` via per-group `DID_{g,l}` building block, with `L_max` parameter | HIGH | Shipped |
+| **2b.** Multi-horizon analytical SE (cohort-recentered plug-in per horizon) | HIGH | Shipped |
+| **2c.** Dynamic placebos `DID^{pl}_l` for pre-trends testing (Web Appendix Section 1.1 of dynamic paper) | HIGH | Shipped |
+| **2d.** Normalized estimator `DID^n_l` (Section 3.2 of dynamic paper) | MEDIUM | Shipped |
+| **2e.** Cost-benefit aggregate `delta` (Section 3.3 of dynamic paper, Lemma 4) | MEDIUM | Shipped |
+| **2f.** Simultaneous (sup-t) confidence bands for event study plots | MEDIUM | Shipped |
+| **2g.** `plot_event_study()` integration; `< 50%`-of-switchers warning for far horizons | MEDIUM | Shipped |
+| **2h.** Parity tests vs `did_multiplegt_dyn` for multi-horizon designs | HIGH | In progress |
 
 ### Phase 3: Covariates, extensions, and tutorial
 

benchmarks/R/generate_dcdh_dynr_test_values.R

@@ -287,6 +287,107 @@ scenarios$hand_calculable_worked_example <- list(
   results = extract_dcdh_l1(res5)
 )
 
+# ---------------------------------------------------------------------------
+# Phase 2: Multi-horizon scenarios (effects > 1)
+# ---------------------------------------------------------------------------
+
+# Helper: extract multi-horizon results from did_multiplegt_dyn output
+extract_dcdh_multi <- function(res, n_effects, n_placebos = 0) {
+  effects <- res$results$Effects
+  if (is.null(effects)) {
+    stop("did_multiplegt_dyn returned no Effects; check the input data")
+  }
+
+  out <- list(effects = list(), placebos = list())
+
+  for (i in seq_len(min(n_effects, nrow(effects)))) {
+    out$effects[[as.character(i)]] <- list(
+      overall_att = as.numeric(effects[i, "Estimate"]),
+      overall_se = as.numeric(effects[i, "SE"]),
+      overall_ci_lo = as.numeric(effects[i, "LB CI"]),
+      overall_ci_hi = as.numeric(effects[i, "UB CI"]),
+      n_switchers = as.numeric(effects[i, "N"])
+    )
+  }
+
+  placebos <- res$results$Placebos
+  if (!is.null(placebos) && n_placebos > 0) {
+    for (i in seq_len(min(n_placebos, nrow(placebos)))) {
+      out$placebos[[as.character(i)]] <- list(
+        effect = as.numeric(placebos[i, "Estimate"]),
+        se = as.numeric(placebos[i, "SE"]),
+        ci_lo = as.numeric(placebos[i, "LB CI"]),
+        ci_hi = as.numeric(placebos[i, "UB CI"])
+      )
+    }
+  }
+
+  out
+}
+
+# Scenario 6: joiners_only multi-horizon (L_max=3, placebo=3)
+# Uses n_periods=8 to give enough room for 3 positive + 3 placebo horizons
+cat("  Scenario 6: joiners_only_multi_horizon\n")
+d6 <- gen_reversible(n_groups = N_GOLDEN, n_periods = 8,
+                     pattern = "joiners_only", seed = 106)
+res6 <- did_multiplegt_dyn(
+  df = d6, outcome = "outcome", group = "group", time = "period",
+  treatment = "treatment", effects = 3, placebo = 3, ci_level = 95
+)
+scenarios$joiners_only_multi_horizon <- list(
+  data = export_data(d6),
+  params = list(pattern = "joiners_only", n_groups = N_GOLDEN, n_periods = 8,
+                seed = 106, effects = 3, placebo = 3, ci_level = 95),
+  results = extract_dcdh_multi(res6, n_effects = 3, n_placebos = 3)
+)
+
+# Scenario 7: leavers_only multi-horizon (L_max=3, placebo=3)
+cat("  Scenario 7: leavers_only_multi_horizon\n")
+d7 <- gen_reversible(n_groups = N_GOLDEN, n_periods = 8,
+                     pattern = "leavers_only", seed = 107)
+res7 <- did_multiplegt_dyn(
+  df = d7, outcome = "outcome", group = "group", time = "period",
+  treatment = "treatment", effects = 3, placebo = 3, ci_level = 95
+)
+scenarios$leavers_only_multi_horizon <- list(
+  data = export_data(d7),
+  params = list(pattern = "leavers_only", n_groups = N_GOLDEN, n_periods = 8,
+                seed = 107, effects = 3, placebo = 3, ci_level = 95),
+  results = extract_dcdh_multi(res7, n_effects = 3, n_placebos = 3)
+)
+
+# Scenario 8: mixed_single_switch multi-horizon (L_max=5, placebo=4)
+# Uses n_periods=10 for far horizons
+cat("  Scenario 8: mixed_single_switch_multi_horizon\n")
+d8 <- gen_reversible(n_groups = N_GOLDEN, n_periods = 10,
+                     pattern = "mixed_single_switch", seed = 108)
+res8 <- did_multiplegt_dyn(
+  df = d8, outcome = "outcome", group = "group", time = "period",
+  treatment = "treatment", effects = 5, placebo = 4, ci_level = 95
+)
+scenarios$mixed_single_switch_multi_horizon <- list(
+  data = export_data(d8),
+  params = list(pattern = "mixed_single_switch", n_groups = N_GOLDEN, n_periods = 10,
+                seed = 108, effects = 5, placebo = 4, ci_level = 95),
+  results = extract_dcdh_multi(res8, n_effects = 5, n_placebos = 4)
+)
+
+# Scenario 9: joiners_only long panel multi-horizon (L_max=5, placebo=5)
+# Uses n_periods=12 and n_groups=80 for thorough coverage
+cat("  Scenario 9: joiners_only_long_multi_horizon\n")
+d9 <- gen_reversible(n_groups = N_GOLDEN, n_periods = 12,
+                     pattern = "joiners_only", seed = 109)
+res9 <- did_multiplegt_dyn(
+  df = d9, outcome = "outcome", group = "group", time = "period",
+  treatment = "treatment", effects = 5, placebo = 5, ci_level = 95
+)
+scenarios$joiners_only_long_multi_horizon <- list(
+  data = export_data(d9),
+  params = list(pattern = "joiners_only", n_groups = N_GOLDEN, n_periods = 12,
+                seed = 109, effects = 5, placebo = 5, ci_level = 95),
+  results = extract_dcdh_multi(res9, n_effects = 5, n_placebos = 5)
+)
+
 # ---------------------------------------------------------------------------
 # Write output
 # ---------------------------------------------------------------------------