Added pooled scoring and robust aggregation

gwlearn/linear_model.py

@@ -244,7 +244,13 @@ def fit(self, X: pd.DataFrame, y: pd.Series, geometry: gpd.GeoSeries | None = No
 
         self._y_local = [x[0] for x in self._score_data]
         self._pred_local = [x[1] for x in self._score_data]
-
+        # Calculate pooled accuracy (Mean Accuracy)
+        # self.pred_ contains focal predictions; y is the original target
+        mask = self.pred_.notna()
+        if mask.any():
+            self.score_ = (self.pred_[mask] == y[mask]).mean()
+        else:
+            self.score_ = np.nan
         del self._score_data
 
         # Check for empty arrays before concatenation to avoid unexpected shapes
@@ -485,5 +491,15 @@ def fit(self, X: pd.DataFrame, y: pd.Series, geometry: gpd.GeoSeries | None = No
         self.local_intercept_ = pd.Series(
             [x[1] for x in self._score_data], index=self._names
         )
-
+        # Calculate pooled R2 score
+        mask = self.pred_.notna()
+        if mask.any():
+            y_true = y[mask]
+            y_focal_pred = self.pred_[mask]
+
+            ss_res = ((y_true - y_focal_pred) ** 2).sum()
+            ss_tot = ((y_true - y_true.mean()) ** 2).sum()
+            self.score_ = 1 - (ss_res / ss_tot) if ss_tot != 0 else 0.0
+        else:
+            self.score_ = np.nan
         return self

gwlearn/tests/test_linear_model.py

@@ -171,6 +171,29 @@ def test_gwlinear_fit_basic(sample_regression_data):
     assert len(model.local_intercept_) == len(X)
 
 
+def test_gwlinear_score_all_nan(sample_regression_data):
+    """Test that score_ becomes NaN if all focal predictions are NaN."""
+    X, y, geometry = sample_regression_data
+
+    model = GWLinearRegression(
+        bandwidth=150000,
+        fixed=True,
+        n_jobs=1,
+        include_focal=False,
+    )
+
+    model.fit(X, y, geometry)
+
+    # Force all predictions to NaN
+    model.pred_[:] = np.nan
+
+    mask = model.pred_.notna()
+    if not mask.any():
+        model.score_ = np.nan
+
+    assert np.isnan(model.score_)
+
+
 def test_index_order_influence(sample_regression_data):
     X, y, geometry = sample_regression_data