diff --git a/tests/conftest.py b/tests/conftest.py
new file mode 100644
index 0000000..5127b32
--- /dev/null
+++ b/tests/conftest.py
@@ -0,0 +1,23 @@
+import os
+
+import pytest
+import pandas as pd
+import numpy as np
+
+
+@pytest.fixture()
+def ovarian_cancer_dataset():
+    dir_path = os.path.dirname(os.path.realpath(__file__))
+
+    data = pd.read_csv(
+        os.path.join(dir_path, "data", "ovarian_cancer.csv"), index_col="idx"
+    )
+
+    # Unfortunately it's unclear how the missing data was imputed in the article
+    # We use median imputation here. As such AUROCs are the same but p-values differ.
+    data.albumin = data.albumin.fillna(np.median(data.albumin.dropna()))
+    data.total_protein = data.total_protein.fillna(
+        np.median(data.total_protein.dropna())
+    )
+
+    return data