init tests

tests/test_pyroc.py

@@ -0,0 +1,81 @@
+from pytest import approx
+from collections import OrderedDict
+
+import numpy as np
+
+import pyroc
+
+
+def test_pyroc_init():
+    roc = pyroc.ROC([0, 0, 1, 1], [0.4, 0.3, 0.8, 0.7])
+    assert roc is not None
+
+
+def test_pyroc_input_parsing(ovarian_cancer_dataset):
+    df = ovarian_cancer_dataset
+
+    expected_values = OrderedDict(
+        [
+            ('albumin', np.asarray([3., 3.2, 3.9])),
+            ('total_protein', np.asarray([5.8, 6.3, 6.8])),
+            # Krebs-Goplerud score
+            ('total_score', np.asarray([0, 5, 7])),
+        ]
+    )
+    n = 3
+
+    # pandas series // pandas dataframe
+    roc = pyroc.ROC(
+        df['outcome'],
+        df.drop('outcome', axis=1),
+    )
+    for i, p in enumerate(expected_values):
+        assert (roc.preds[p][:n] == expected_values[p][:n]).all()
+
+    # from now on we only pass target as numpy array
+    target = df['outcome'].values
+    df = df.drop('outcome', axis=1)
+
+    # numpy arrays
+    preds = df.values
+    roc = pyroc.ROC(
+        target,
+        preds,
+    )
+    for i, p in enumerate(expected_values):
+        # since roc wasn't provided predictor labels,
+        # keys are integers (i)
+        assert (roc.preds[i][:n] == expected_values[p][:n]).all()
+
+    # list
+    preds = [df[c].values for c in df.columns]
+    roc = pyroc.ROC(
+        target,
+        preds,
+    )
+    for i, p in enumerate(expected_values):
+        # since roc wasn't provided predictor labels,
+        # keys are integers (i)
+        assert (roc.preds[i][:n] == expected_values[p][:n]).all()
+
+    # dict
+    preds = {c: df[c].values for c in df.columns}
+    roc = pyroc.ROC(
+        target,
+        preds,
+    )
+    for i, p in enumerate(expected_values):
+        assert p in roc.preds.keys()
+        assert (roc.preds[p][:n] == expected_values[p][:n]).all()
+
+
+def test_pyroc_compare(ovarian_cancer_dataset):
+    roc = pyroc.ROC(
+        ovarian_cancer_dataset['outcome'],
+        ovarian_cancer_dataset.drop('outcome', axis=1),
+    )
+    p, ci = roc.compare(np.array([[1, -1, 0], [1, 0, -1]]))
+
+    assert p == approx(0.42291256338064165)
+    assert ci[0] == approx(0.05063562)
+    assert ci[1] == approx(7.37775891)