Fix CI review Round 5: non-binary metadata, suppress binary decomposition

P1: For non-binary L_max >= 1, use per-group N_l as n_switcher_cells
    (not binary N_S=0), count non-baseline obs as n_treated_obs, and
    suppress joiner/leaver decomposition (binary concept). Ensures
    DID_1 label pairs with consistent per-group metadata.
P2: Add test_nonbinary_lmax1_renderer_contract asserting DID_1 label
    in __repr__/to_dataframe, n_switcher_cells > 0, joiners unavailable.
P3: Update to_dataframe() docstring for L_max >= 1 contract.

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

Author: igerber

diff_diff/chaisemartin_dhaultfoeuille.py

@@ -1831,28 +1831,53 @@ def fit(
             twfe_sigma_fe = twfe_diagnostic_payload.sigma_fe
             twfe_beta_fe = twfe_diagnostic_payload.beta_fe
 
+        # When L_max >= 1 on non-binary data, the binary-only metadata
+        # (N_S, joiner/leaver counts, n_treated_obs) doesn't match the
+        # per-group DID_1 estimand. Use per-group metadata instead and
+        # suppress the joiner/leaver decomposition.
+        effective_N_S = N_S
+        effective_n_treated = n_treated_obs_post
+        effective_joiners_available = joiners_available
+        effective_leavers_available = leavers_available
+        if (
+            not is_binary
+            and L_max is not None
+            and L_max >= 1
+            and multi_horizon_dids is not None
+            and 1 in multi_horizon_dids
+        ):
+            # Use horizon-1 eligible switcher count as the effective N_S
+            effective_N_S = multi_horizon_dids[1]["N_l"]
+            # Count all observations where treatment differs from baseline
+            effective_n_treated = int(
+                N_mat[D_mat != D_mat[:, 0:1]].sum()
+            ) if D_mat.shape[1] > 1 else 0
+            # Suppress joiner/leaver decomposition for non-binary
+            effective_joiners_available = False
+            effective_leavers_available = False
+
         results = ChaisemartinDHaultfoeuilleResults(
             overall_att=effective_overall_att,
             overall_se=effective_overall_se,
             overall_t_stat=effective_overall_t,
             overall_p_value=effective_overall_p,
             overall_conf_int=effective_overall_ci,
-            joiners_att=joiners_att,
-            joiners_se=joiners_se,
-            joiners_t_stat=joiners_t,
-            joiners_p_value=joiners_p,
-            joiners_conf_int=joiners_ci,
-            n_joiner_cells=n_joiner_cells,
-            n_joiner_obs=n_joiner_obs,
-            joiners_available=joiners_available,
-            leavers_att=leavers_att,
-            leavers_se=leavers_se,
-            leavers_t_stat=leavers_t,
-            leavers_p_value=leavers_p,
-            leavers_conf_int=leavers_ci,
-            n_leaver_cells=n_leaver_cells,
-            n_leaver_obs=n_leaver_obs,
-            leavers_available=leavers_available,
+            joiners_att=joiners_att if effective_joiners_available else float("nan"),
+            joiners_se=joiners_se if effective_joiners_available else float("nan"),
+            joiners_t_stat=joiners_t if effective_joiners_available else float("nan"),
+            joiners_p_value=joiners_p if effective_joiners_available else float("nan"),
+            joiners_conf_int=joiners_ci if effective_joiners_available else (float("nan"), float("nan")),
+            n_joiner_cells=n_joiner_cells if effective_joiners_available else 0,
+            n_joiner_obs=n_joiner_obs if effective_joiners_available else 0,
+            joiners_available=effective_joiners_available,
+            leavers_att=leavers_att if effective_leavers_available else float("nan"),
+            leavers_se=leavers_se if effective_leavers_available else float("nan"),
+            leavers_t_stat=leavers_t if effective_leavers_available else float("nan"),
+            leavers_p_value=leavers_p if effective_leavers_available else float("nan"),
+            leavers_conf_int=leavers_ci if effective_leavers_available else (float("nan"), float("nan")),
+            n_leaver_cells=n_leaver_cells if effective_leavers_available else 0,
+            n_leaver_obs=n_leaver_obs if effective_leavers_available else 0,
+            leavers_available=effective_leavers_available,
             placebo_effect=placebo_effect,
             placebo_se=placebo_se,
             placebo_t_stat=placebo_t,
@@ -1863,8 +1888,8 @@ def fit(
             groups=all_groups,
             time_periods=all_periods,
             n_obs=n_obs_post,
-            n_treated_obs=n_treated_obs_post,
-            n_switcher_cells=N_S,
+            n_treated_obs=effective_n_treated,
+            n_switcher_cells=effective_N_S,
             n_cohorts=n_cohorts,
             n_groups_dropped_crossers=n_groups_dropped_crossers,
             n_groups_dropped_singleton_baseline=n_groups_dropped_singleton_baseline,

diff_diff/chaisemartin_dhaultfoeuille_results.py

@@ -778,10 +778,11 @@ def to_dataframe(self, level: str = "overall") -> pd.DataFrame:
         level : str, default="overall"
             One of:
 
-            - ``"overall"``: single-row table with the overall ``DID_M``
-              point estimate, SE, t-stat, p-value, CI bounds.
-            - ``"joiners_leavers"``: three rows for ``DID_M``, ``DID_+``,
-              and ``DID_-``.
+            - ``"overall"``: single-row table with the overall estimand
+              (``DID_M`` when ``L_max=None``, ``DID_1`` when ``L_max=1``,
+              ``delta`` when ``L_max >= 2``).
+            - ``"joiners_leavers"``: up to three rows for the overall,
+              ``DID_+``, and ``DID_-`` (binary panels only).
             - ``"per_period"``: one row per time period with
               ``did_plus_t``, ``did_minus_t``, switching cell counts, and
               the A11-zeroed flags.

tests/test_chaisemartin_dhaultfoeuille.py

@@ -2422,6 +2422,39 @@ def test_nonbinary_bootstrap(self, ci_params):
         assert r.overall_se == es1["se"]
         assert r.overall_p_value == es1["p_value"]
 
+    def test_nonbinary_lmax1_renderer_contract(self):
+        """Non-binary L_max=1: summary/to_dataframe use DID_1 label and
+        suppress binary-only joiner/leaver decomposition."""
+        np.random.seed(77)
+        rows = []
+        for g in range(20):
+            for t in range(6):
+                d = 0 if t < 3 else 2
+                y = 10 + t + d + np.random.randn() * 0.3
+                rows.append({"group": g, "period": t, "treatment": d, "outcome": y})
+        for g in range(20, 40):
+            for t in range(6):
+                y = 10 + t + np.random.randn() * 0.3
+                rows.append({"group": g, "period": t, "treatment": 0, "outcome": y})
+        df = pd.DataFrame(rows)
+        est = ChaisemartinDHaultfoeuille(twfe_diagnostic=False)
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore")
+            r = est.fit(
+                df, outcome="outcome", group="group", time="period",
+                treatment="treatment", L_max=1,
+            )
+        # __repr__ should say DID_1
+        assert "DID_1" in repr(r)
+        # to_dataframe("overall") should label as DID_1
+        df_overall = r.to_dataframe("overall")
+        assert df_overall.iloc[0]["estimand"] == "DID_1"
+        # n_switcher_cells should be > 0 (from per-group path)
+        assert r.n_switcher_cells > 0
+        # Joiners/leavers unavailable for non-binary
+        assert r.joiners_available is False
+        assert r.leavers_available is False
+
     def test_twfe_diagnostic_skipped_nonbinary(self):
         """TWFE diagnostic should be skipped (with warning) for non-binary."""
         np.random.seed(77)