Merge pull request #389 from CITCOM-project/jmafoster1/hotfix

Fixed numpy linalg error

Author: jmafoster1

causal_testing/discovery/abstract_discovery.py

@@ -10,6 +10,7 @@
 from itertools import permutations
 
 import networkx as nx
+import numpy as np
 import pandas as pd
 import rustworkx as rx
 
@@ -207,28 +208,27 @@ def evaluate_tests(self, causal_dag: CausalDAG) -> pd.DataFrame:
 
         results = []
 
-        # We use "silent=False" here to allow for inestimable edges, but it'd be good to have a more stringent
-        # error catching strategy to catch "genuine" problems (e.g. to do with the structure of the data)
-        for test_case, result in zip(ctf.test_cases, ctf.run_tests(silent=False)):
-            if result.effect_estimate is None:
+        for test_case, result in zip(ctf.test_cases, ctf.test_cases):
+            try:
+                result = test_case.execute_test()
                 results.append(
                     {
-                        "result": TestResult.INESTIMABLE,
+                        "result": (
+                            TestResult.PASS if test_case.expected_causal_effect.apply(result) else TestResult.FAIL
+                        ),
                         "expected_effect": test_case.expected_causal_effect.__class__.__name__,
                         "treatment": test_case.base_test_case.treatment_variable.name,
                         "outcome": test_case.base_test_case.outcome_variable.name,
+                        "effect": effect_direction(result),
                     }
                 )
-            else:
+            except np.linalg.LinAlgError:
                 results.append(
                     {
-                        "result": (
-                            TestResult.PASS if test_case.expected_causal_effect.apply(result) else TestResult.FAIL
-                        ),
+                        "result": TestResult.INESTIMABLE,
                         "expected_effect": test_case.expected_causal_effect.__class__.__name__,
                         "treatment": test_case.base_test_case.treatment_variable.name,
                         "outcome": test_case.base_test_case.outcome_variable.name,
-                        "effect": effect_direction(result),
                     }
                 )
 

tests/discovery_tests/test_abstract_discovery.py

@@ -176,6 +176,105 @@ def test_write_dot_invalid_independence_outcome(self):
         with self.assertRaises(ValueError):
             abstract_discovery.write_dot(dag, "dag.dot")
 
+    def test_evaluate_tests_invalid_datatype(self):
+        scarf_df = pd.read_csv("tests/resources/data/scarf_data.csv")
+        scarf_df["completed"] = pd.to_datetime(["2026-01-01" for _ in range(len(scarf_df))], format="%Y-%m-%d")
+
+        dag = CausalDAG()
+        dag.add_nodes_from(scarf_df.columns)
+        dag.add_edges_from([("length_in", "completed"), ("large_gauge", "completed")])
+
+        abstract_discovery = AbstractDiscovery(scarf_df)
+        with self.assertRaises(ValueError):
+            abstract_discovery.evaluate_tests(dag)
+
+    def test_evaluate_tests_inestimable(self):
+        scarf_df = pd.read_csv("tests/resources/data/scarf_data.csv")
+        scarf_df["completed"] = scarf_df["completed"].astype(bool)
+        scarf_df = scarf_df.loc[(scarf_df["completed"]) & (scarf_df["color"] != "grey")]
+
+        dag = CausalDAG()
+        dag.add_nodes_from(scarf_df.columns)
+        dag.add_edges_from([("length_in", "completed"), ("large_gauge", "completed")])
+
+        abstract_discovery = AbstractDiscovery(scarf_df)
+        test_results = abstract_discovery.evaluate_tests(dag)
+        expected_results = pd.DataFrame(
+            [
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "length_in",
+                    "outcome": "large_gauge",
+                    "effect": "negative",
+                },
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "large_gauge",
+                    "outcome": "length_in",
+                    "effect": "negative",
+                },
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "length_in",
+                    "outcome": "color",
+                    "effect": None,
+                },
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "color",
+                    "outcome": "length_in",
+                    "effect": None,
+                },
+                {
+                    "result": TestResult.FAIL,
+                    "expected_effect": "SomeEffect",
+                    "treatment": "length_in",
+                    "outcome": "completed",
+                    "effect": "positive",
+                },
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "large_gauge",
+                    "outcome": "color",
+                    "effect": None,
+                },
+                {
+                    "result": TestResult.PASS,
+                    "expected_effect": "NoEffect",
+                    "treatment": "color",
+                    "outcome": "large_gauge",
+                    "effect": None,
+                },
+                {
+                    "result": TestResult.FAIL,
+                    "expected_effect": "SomeEffect",
+                    "treatment": "large_gauge",
+                    "outcome": "completed",
+                    "effect": "negative",
+                },
+                {
+                    "result": TestResult.INESTIMABLE,
+                    "expected_effect": "NoEffect",
+                    "treatment": "color",
+                    "outcome": "completed",
+                    "effect": None,
+                },
+                {
+                    "result": TestResult.INESTIMABLE,
+                    "expected_effect": "NoEffect",
+                    "treatment": "completed",
+                    "outcome": "color",
+                    "effect": None,
+                },
+            ]
+        )
+        pd.testing.assert_frame_equal(test_results, expected_results)
+
     def test_evaluate_tests(self):
         scarf_df = pd.read_csv("tests/resources/data/scarf_data.csv")