fix(pearson_coefficient): ensure that it works the same as the python…

… correlation function + tests

causy/generators.py

@@ -66,6 +66,9 @@ def __init__(
         if shuffle_combinations is not None:
             self.shuffle_combinations = shuffle_combinations
 
+        if chunked is not None:
+            self.chunked = chunked
+
     def serialize(self):
         result = super().serialize()
         result["params"]["shuffle_combinations"] = self.shuffle_combinations

causy/utils.py

@@ -64,6 +64,6 @@ def retrieve_edges(graph):
 
 def pearson_correlation(x, y):
     cov_xy = torch.mean((x - x.mean()) * (y - y.mean()))
-    std_x = x.std()
-    std_y = y.std()
+    std_x = x.std(unbiased=False)
+    std_y = y.std(unbiased=False)
     return cov_xy / (std_x * std_y)

tests/test_utils.py

@@ -1,8 +1,47 @@
 import unittest
+from statistics import correlation
+
+from causy.independence_tests import CalculateCorrelations
+from causy.utils import (
+    pearson_correlation,
+    serialize_module_name,
+    load_pipeline_artefact_by_definition,
+    load_pipeline_steps_by_definition,
+)
+import torch
 
 
 class UtilsTestCase(unittest.TestCase):
-    pass
+    def test_pearson_correlation(self):
+        self.assertEqual(
+            pearson_correlation(
+                torch.tensor([1, 2, 3], dtype=torch.float64),
+                torch.tensor([1, 2, 3], dtype=torch.float64),
+            ).item(),
+            1,
+        )
+        self.assertEqual(
+            pearson_correlation(
+                torch.tensor([1, 2, 3], dtype=torch.float64),
+                torch.tensor([3, 2, 1], dtype=torch.float64),
+            ).item(),
+            -1,
+        )
+
+    def test_serialize_module_name(self):
+        self.assertEqual(serialize_module_name(self), "tests.test_utils.UtilsTestCase")
+
+    def test_load_pipeline_artefact_by_definition(self):
+        step = {"name": "causy.independence_tests.CalculateCorrelations"}
+        self.assertIsInstance(
+            load_pipeline_artefact_by_definition(step), CalculateCorrelations
+        )
+
+    def load_pipeline_steps_by_definition(self):
+        steps = [{"name": "causy.independence_tests.CalculateCorrelations"}]
+        self.assertIsInstance(
+            load_pipeline_steps_by_definition(steps)[0], CalculateCorrelations
+        )
 
 
 if __name__ == "__main__":