Address PR #192 review (Round 5): balance_e empty warning, inline comments for repeat false positives

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: igerber

diff_diff/efficient_did.py

@@ -319,6 +319,11 @@ def fit(
             else:
                 effective_p1_col = period_1_col
 
+            # Estimate all (g, t) cells including pre-treatment. Under PT-Post,
+            # pre-treatment cells serve as placebo/pre-trend diagnostics, matching
+            # the CallawaySantAnna implementation. Users filter to t >= g for
+            # post-treatment effects; pre-treatment cells are clearly labeled by
+            # their (g, t) coordinates in the results object.
             for t in time_periods:
                 # Skip period_1 — it's the universal reference baseline,
                 # not a target period
@@ -713,6 +718,14 @@ def _aggregate_event_study(
                     balanced[e].append(((g, t), data["effect"], cohort_fractions.get(g, 0.0)))
             effects_by_e = balanced
 
+        if balance_e is not None and not effects_by_e:
+            warnings.warn(
+                f"balance_e={balance_e}: no cohort has a finite effect at the "
+                "anchor horizon. Event study will be empty.",
+                UserWarning,
+                stacklevel=2,
+            )
+
         result: Dict[int, Dict[str, Any]] = {}
         for e, elist in sorted(effects_by_e.items()):
             gt_pairs = [x[0] for x in elist]

diff_diff/efficient_did_bootstrap.py

@@ -116,7 +116,12 @@ def _run_multiplier_bootstrap(
         )
         post_indices = np.where(post_mask)[0]
 
-        # Overall ATT aggregation weights (cohort-size)
+        # Overall ATT: fixed-weight re-aggregation of perturbed cell ATTs.
+        # This matches CallawaySantAnna._run_multiplier_bootstrap
+        # (staggered_bootstrap.py:281). The analytical path includes a WIF
+        # correction; bootstrap captures sampling variability through per-cell
+        # EIF perturbation without re-estimating weights — this is standard
+        # in both this library's CS implementation and the R did package.
         skip_overall = len(post_indices) == 0
         if skip_overall:
             bootstrap_overall = np.full(self.n_bootstrap, np.nan)
@@ -129,7 +134,8 @@ def _run_multiplier_bootstrap(
             with np.errstate(divide="ignore", invalid="ignore", over="ignore"):
                 bootstrap_overall = bootstrap_atts[:, post_indices] @ agg_w
 
-        # Event study aggregation
+        # Event study: fixed-weight re-aggregation (same pattern as overall).
+        # See note above re: WIF — analytical WIF is not needed in bootstrap.
         bootstrap_event_study = None
         event_study_info = None
         if aggregate in ("event_study", "all"):
@@ -261,6 +267,14 @@ def _prepare_es_agg_boot(
                     balanced[e].append((j, original_atts[j], cohort_fractions.get(g, 0.0)))
             effects_by_e = balanced
 
+        if balance_e is not None and not effects_by_e:
+            warnings.warn(
+                f"balance_e={balance_e}: no cohort has a finite effect at the "
+                "anchor horizon. Event study will be empty.",
+                UserWarning,
+                stacklevel=2,
+            )
+
         result = {}
         for e, elist in effects_by_e.items():
             indices = np.array([x[0] for x in elist])

tests/test_efficient_did.py

@@ -719,6 +719,22 @@ def test_balance_e_nan_anchor_filters_group(self):
                 f"Group 5 (NaN at anchor) should be excluded at e={e}, " f"got groups {groups_in_e}"
             )
 
+    def test_balance_e_empty_warns(self):
+        """When no cohort survives the anchor horizon, warn the user."""
+        edid = EfficientDiD()
+        edid.anticipation = 0
+
+        # All effects are NaN at e=0
+        gt_pairs = [(3.0, 3), (3.0, 4), (5.0, 5), (5.0, 6)]
+        original_atts = np.array([np.nan, 1.5, np.nan, 0.8])
+        cohort_fractions = {3.0: 0.4, 5.0: 0.3}
+
+        with pytest.warns(UserWarning, match="no cohort has a finite effect"):
+            result = edid._prepare_es_agg_boot(
+                gt_pairs, original_atts, cohort_fractions, balance_e=0
+            )
+        assert result == {}
+
 
 # =============================================================================
 # Tier 3: Bootstrap