Misimplemented e val. Re implemented

Author: rsomers1998

causal_testing/testing/validation.py

@@ -24,33 +24,36 @@ def estimate_robustness(self, model: RegressionResultsWrapper, q=1, alpha=1):
 
         return rv
 
-    def estimate_e_value(
-        self, risk_ratio, confidence_intervals: tuple[float, float]
-    ) -> tuple[float, tuple[float, float]]:
+    def estimate_e_value(self, risk_ratio: float) -> float:
         """Calculate the E value from a risk ratio. This allow
         the user to identify how large a risk an unidentified confounding
         variable would need to be to nullify the causal relationship
         under test."""
 
         if risk_ratio >= 1:
-            e = risk_ratio + math.sqrt(risk_ratio * (risk_ratio - 1))
+            return risk_ratio + math.sqrt(risk_ratio * (risk_ratio - 1))
 
-            lower_limit = confidence_intervals[0]
-            if lower_limit <= 1:
-                lower_limit = 1
-            else:
-                lower_limit = lower_limit + math.sqrt(lower_limit * (lower_limit - 1))
+        risk_ratio_prime = 1 / risk_ratio
+        return risk_ratio_prime + math.sqrt(risk_ratio_prime * (risk_ratio_prime - 1))
 
-            return (e, (lower_limit, 1))
+    def estimate_e_value_using_ci(self, risk_ratio: float, confidence_intervals: tuple[float, float]) -> float:
+        """Calculate the E value from a risk ratio and it's confidence intervals.
+        This allow the user to identify how large a risk an unidentified
+        confounding variable would need to be to nullify the causal relationship
+        under test."""
 
-        risk_ratio_prime = 1 / risk_ratio
-        e = risk_ratio_prime + math.sqrt(risk_ratio_prime * (risk_ratio_prime - 1))
+        if risk_ratio >= 1:
+            lower_limit = confidence_intervals[0]
+            e = 1
+            if lower_limit > 1:
+                e = lower_limit + math.sqrt(lower_limit * (lower_limit - 1))
+
+            return e
 
         upper_limit = confidence_intervals[1]
-        if upper_limit >= 1:
-            upper_limit = 1
-        else:
+        e = 1
+        if upper_limit < 1:
             upper_limit_prime = 1 / upper_limit
-            upper_limit = upper_limit_prime + math.sqrt(upper_limit_prime * (upper_limit_prime - 1))
+            e = upper_limit_prime + math.sqrt(upper_limit_prime * (upper_limit_prime - 1))
 
-        return (e, (1, upper_limit))
+        return e

tests/testing_tests/test_causal_test_outcome.py

@@ -179,31 +179,21 @@ def test_someEffect_fail(self):
         )
 
     def test_positive_risk_ratio_e_value(self):
-        test_value = TestValue("risk_ratio", 1.5)
-        ctr = CausalTestResult(
-            estimator=self.estimator,
-            test_value=test_value,
-            confidence_intervals=[1.2, 1.8],
-            effect_modifier_configuration=None,
-        )
-
         cv = CausalValidator()
-        e_value, e_confidence_intervals = cv.estimate_e_value(ctr.test_value.value, ctr.confidence_intervals)
+        e_value = cv.estimate_e_value(1.5)
         self.assertEqual(round(e_value, 4), 2.366)
-        self.assertEqual(round(e_confidence_intervals[0], 4), 1.6899)
-        self.assertEqual(e_confidence_intervals[1], 1)
 
-    def test_negative_risk_ratio_e_value(self):
-        test_value = TestValue("risk_ratio", 0.8)
-        ctr = CausalTestResult(
-            estimator=self.estimator,
-            test_value=test_value,
-            confidence_intervals=[0.2, 0.9],
-            effect_modifier_configuration=None,
-        )
+    def test_positive_risk_ratio_e_value_using_ci(self):
+        cv = CausalValidator()
+        e_value = cv.estimate_e_value_using_ci(1.5, [1.2, 1.8])
+        self.assertEqual(round(e_value, 4), 1.6899)
 
+    def test_negative_risk_ratio_e_value(self):
         cv = CausalValidator()
-        e_value, e_confidence_intervals = cv.estimate_e_value(ctr.test_value.value, ctr.confidence_intervals)
+        e_value = cv.estimate_e_value(0.8)
         self.assertEqual(round(e_value, 4), 1.809)
-        self.assertEqual(e_confidence_intervals[0], 1)
-        self.assertEqual(round(e_confidence_intervals[1], 4), 1.4625)
+
+    def test_negative_risk_ratio_e_value_using_ci(self):
+        cv = CausalValidator()
+        e_value = cv.estimate_e_value_using_ci(0.8, [0.2, 0.9])
+        self.assertEqual(round(e_value, 4), 1.4625)