lin reg: mention missing & superfluous predictors (#359)

Author: mathause

Diff for: mesmer/stats/_linear_regression.py

@@ -80,8 +80,15 @@ def predict(
         required_predictors = set(params.data_vars) - non_predictor_vars - exclude
         available_predictors = set(predictors.keys())
 
-        if required_predictors != available_predictors:
-            raise ValueError("Missing or superfluous predictors.")
+        if required_predictors - available_predictors:
+            missing = sorted(required_predictors - available_predictors)
+            missing = "', '".join(missing)
+            raise ValueError(f"Missing predictors: '{missing}'")
+
+        if available_predictors - required_predictors:
+            superfluous = sorted(available_predictors - required_predictors)
+            superfluous = "', '".join(superfluous)
+            raise ValueError(f"Superfluous predictors: '{superfluous}'")
 
         if "intercept" in exclude:
             prediction = xr.zeros_like(params.intercept)

Diff for: tests/unit/test_linear_regression.py

@@ -36,7 +36,7 @@ def LinearRegression_fit_wrapper(*args, **kwargs):
 # TEST LinearRegression class
 
 
-def test_LR_params():
+def test_lr_params():
 
     lr = mesmer.stats.LinearRegression()
 
@@ -80,20 +80,14 @@ def test_LR_params():
 
 
 @pytest.mark.parametrize("as_2D", [True, False])
-def test_LR_predict(as_2D):
+def test_lr_predict(as_2D):
     lr = mesmer.stats.LinearRegression()
 
     params = xr.Dataset(
         data_vars={"intercept": ("x", [5]), "fit_intercept": True, "tas": ("x", [3])}
     )
     lr.params = params if as_2D else params.squeeze()
 
-    with pytest.raises(ValueError, match="Missing or superfluous predictors"):
-        lr.predict({})
-
-    with pytest.raises(ValueError, match="Missing or superfluous predictors"):
-        lr.predict({"tas": None, "something else": None})
-
     tas = xr.DataArray([0, 1, 2], dims="time")
 
     result = lr.predict({"tas": tas})
@@ -103,6 +97,32 @@ def test_LR_predict(as_2D):
     xr.testing.assert_equal(result, expected)
 
 
+def test_lr_predict_missing_superfluous():
+    lr = mesmer.stats.LinearRegression()
+
+    params = xr.Dataset(
+        data_vars={
+            "intercept": ("x", [5]),
+            "fit_intercept": True,
+            "tas": ("x", [3]),
+            "tas2": ("x", [1]),
+        }
+    )
+    lr.params = params
+
+    with pytest.raises(ValueError, match="Missing predictors: 'tas', 'tas2'"):
+        lr.predict({})
+
+    with pytest.raises(ValueError, match="Missing predictors: 'tas'"):
+        lr.predict({"tas2": None})
+
+    with pytest.raises(ValueError, match="Superfluous predictors: 'something else'"):
+        lr.predict({"tas": None, "tas2": None, "something else": None})
+
+    with pytest.raises(ValueError, match="Superfluous predictors: 'bar', 'foo'"):
+        lr.predict({"tas": None, "tas2": None, "foo": None, "bar": None})
+
+
 @pytest.mark.parametrize("as_2D", [True, False])
 def test_lr_predict_exclude(as_2D):
     lr = mesmer.stats.LinearRegression()