Fix epv_summary column schema on empty results, update test comments

Ensure epv_summary(show_all=False) returns DataFrame with correct
column schema even when no entries have low EPV, across all three
results classes. Fix remaining test comments to use intercept-excluded
EPV arithmetic.

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

Author: igerber

diff_diff/staggered_results.py

@@ -310,7 +310,8 @@ def epv_summary(self, show_all: bool = False) -> pd.DataFrame:
                         "is_low": diag.get("is_low", False),
                     }
                 )
-        return pd.DataFrame(rows)
+        cols = ["group", "time", "epv", "n_events", "n_params", "is_low"]
+        return pd.DataFrame(rows, columns=cols) if rows else pd.DataFrame(columns=cols)
 
     def print_summary(self, alpha: Optional[float] = None) -> None:
         """Print summary to stdout."""

diff_diff/staggered_triple_diff_results.py

@@ -286,7 +286,8 @@ def epv_summary(self, show_all: bool = False) -> pd.DataFrame:
                         "is_low": diag.get("is_low", False),
                     }
                 )
-        return pd.DataFrame(rows)
+        cols = ["group", "time", "epv", "n_events", "n_params", "is_low"]
+        return pd.DataFrame(rows, columns=cols) if rows else pd.DataFrame(columns=cols)
 
     def to_dataframe(self, level: str = "group_time") -> pd.DataFrame:
         """

diff_diff/triple_diff.py

@@ -335,7 +335,8 @@ def epv_summary(self, show_all: bool = False) -> pd.DataFrame:
                         "is_low": diag.get("is_low", False),
                     }
                 )
-        return pd.DataFrame(rows)
+        cols = ["subgroup", "epv", "n_events", "n_params", "is_low"]
+        return pd.DataFrame(rows, columns=cols) if rows else pd.DataFrame(columns=cols)
 
 
 # =============================================================================

tests/test_linalg.py

@@ -1728,11 +1728,11 @@ def test_epv_threshold_configurable(self):
         # 15 events, 2 predictor variables → EPV = 7.5
         y = np.concatenate([np.ones(15), np.zeros(n - 15)])
 
-        # Default threshold 10 → should warn (EPV=5)
+        # Default threshold 10 → should warn (EPV=7.5 < 10)
         with pytest.warns(UserWarning, match="Low Events Per Variable"):
             solve_logit(X, y, epv_threshold=10)
 
-        # Threshold 3 → should not warn (EPV=5 >= 3)
+        # Threshold 3 → should not warn (EPV=7.5 >= 3)
         import warnings
 
         with warnings.catch_warnings(record=True) as w: