Audit combined stratification isolation

tests/test_stratification_isolation.py

@@ -0,0 +1,108 @@
+import numpy as np
+from polars.testing import assert_frame_equal
+
+from rtichoke import prepare_performance_data, prepare_performance_data_times
+
+
+def _sort_binary(df):
+    return df.sort(["reference_group", "stratified_by", "chosen_cutoff"])
+
+
+def _sort_times(df):
+    return df.sort(
+        [
+            "fixed_time_horizon",
+            "censoring_heuristic",
+            "competing_heuristic",
+            "stratified_by",
+            "chosen_cutoff",
+            "reference_group",
+        ]
+    )
+
+
+def test_binary_combined_stratification_preserves_each_component():
+    probs = {
+        "group_a": np.array([0.05, 0.15, 0.35, 0.55, 0.75, 0.95]),
+        "group_b": np.array([0.10, 0.25, 0.40, 0.60, 0.80, 0.90]),
+    }
+    reals = {
+        "group_a": np.array([0, 0, 1, 0, 1, 1]),
+        "group_b": np.array([0, 1, 0, 1, 1, 1]),
+    }
+
+    combined = prepare_performance_data(
+        probs=probs,
+        reals=reals,
+        stratified_by=["probability_threshold", "ppcr"],
+        by=0.25,
+    )
+
+    for stratification in ["probability_threshold", "ppcr"]:
+        isolated = prepare_performance_data(
+            probs=probs,
+            reals=reals,
+            stratified_by=[stratification],
+            by=0.25,
+        )
+        from_combined = combined.filter(
+            combined["stratified_by"] == stratification
+        )
+        assert_frame_equal(
+            _sort_binary(from_combined),
+            _sort_binary(isolated),
+            check_row_order=True,
+            check_column_order=True,
+        )
+
+
+def test_time_combined_stratification_preserves_each_component_across_groups_and_horizons():
+    probs = {
+        "group_a": np.array([0.05, 0.15, 0.35, 0.55, 0.75, 0.95]),
+        "group_b": np.array([0.10, 0.25, 0.40, 0.60, 0.80, 0.90]),
+    }
+    reals = {
+        "group_a": np.array([0, 1, 0, 1, 2, 1]),
+        "group_b": np.array([0, 0, 1, 2, 1, 1]),
+    }
+    times = {
+        "group_a": np.array([0.5, 1.0, 1.5, 2.0, 2.5, 3.0]),
+        "group_b": np.array([0.4, 1.2, 1.6, 2.2, 2.6, 3.2]),
+    }
+    horizons = [1.5, 2.5]
+    heuristics = [
+        {
+            "censoring_heuristic": "adjusted",
+            "competing_heuristic": "adjusted_as_negative",
+        }
+    ]
+
+    combined = prepare_performance_data_times(
+        probs=probs,
+        reals=reals,
+        times=times,
+        fixed_time_horizons=horizons,
+        heuristics_sets=heuristics,
+        stratified_by=["probability_threshold", "ppcr"],
+        by=0.25,
+    )
+
+    for stratification in ["probability_threshold", "ppcr"]:
+        isolated = prepare_performance_data_times(
+            probs=probs,
+            reals=reals,
+            times=times,
+            fixed_time_horizons=horizons,
+            heuristics_sets=heuristics,
+            stratified_by=[stratification],
+            by=0.25,
+        )
+        from_combined = combined.filter(
+            combined["stratified_by"] == stratification
+        )
+        assert_frame_equal(
+            _sort_times(from_combined),
+            _sort_times(isolated),
+            check_row_order=True,
+            check_column_order=True,
+        )