Fix unit test

Signed-off-by: Adam Li <adam2392@gmail.com>

tests/test_api.py

@@ -3,14 +3,14 @@
 
 from pywhy_stats import Methods, PValueResult, independence_test
 
+rng = np.random.default_rng(0)
+
 
 def test_independence_test():
     # Replace these arrays with appropriate test data for your specific case
-    X = np.random.standard_normal(size=(20, 3))  # 10 samples, 3 features for X
-    Y = np.random.standard_normal(size=(20, 3))  # 10 samples, 3 features for Y
-    condition_on = np.random.standard_normal(
-        size=(20, 3)
-    )  # 10 samples, 3 features for condition_on
+    X = rng.standard_normal(size=(20, 3))  # 10 samples, 3 features for X
+    Y = rng.standard_normal(size=(20, 3))  # 10 samples, 3 features for Y
+    condition_on = rng.standard_normal(size=(20, 3))  # 10 samples, 3 features for condition_on
 
     # Test case for unconditional independence test
     result = independence_test(X, Y, method=Methods.FISHERZ)
@@ -28,9 +28,11 @@ def test_independence_test():
         independence_test(X, Y, method="unsupported_method")
 
     # Test invalid data is caught with testing for Gaussianity
-    n_samples = 2000
-    X = np.random.uniform(low=0.0, high=10, size=(n_samples, 3))  # 10 samples, 3 features for X
-    Y = np.random.uniform(size=(n_samples, 3))  # 10 samples, 3 features for X
-    condition_on = np.random.uniform(size=(n_samples, 3))  # 10 samples, 3 features for condition_on
+    n_samples = 1000
+    X1 = rng.uniform(low=-5, high=0, size=(n_samples // 2, 3))  # 10 samples, 3 features for X
+    X2 = rng.normal(loc=10, scale=1, size=(n_samples // 2, 3))  # 10 samples, 3 features for X
+    X = np.concatenate((X1, X2), axis=0)
+    Y = rng.uniform(size=(n_samples, 3))  # 10 samples, 3 features for X
+    condition_on = rng.uniform(size=(n_samples, 3))  # 10 samples, 3 features for condition_on
     with pytest.warns(UserWarning, match="The provided data does not seem to be Gaussian"):
         independence_test(X, Y, condition_on, method=Methods.FISHERZ)