Add data drift Anderson-Darling test

Author: jaime-cespedes-sisniega

README.md

@@ -290,9 +290,15 @@ The currently implemented detectors are listed in the following table.
     <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.1007/978-3-540-75488-6_27">Nishida and Yamauchi (2007)</a></td>
   </tr>
   <tr>
-    <td rowspan="15" style="text-align: center; border: 1px solid grey; padding: 8px;">Data drift</td>
-    <td rowspan="13" style="text-align: center; border: 1px solid grey; padding: 8px;">Batch</td>
-    <td rowspan="8" style="text-align: center; border: 1px solid grey; padding: 8px;">Distance based</td>
+    <td rowspan="16" style="text-align: center; border: 1px solid grey; padding: 8px;">Data drift</td>
+    <td rowspan="14" style="text-align: center; border: 1px solid grey; padding: 8px;">Batch</td>
+    <td rowspan="9" style="text-align: center; border: 1px solid grey; padding: 8px;">Distance based</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">N</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;">Anderson-Darling test</td>
+    <td style="text-align: center; border: 1px solid grey; padding: 8px;"><a href="https://doi.org/10.2307/2288805">Scholz and Stephens (1987)</a></td>
+  </tr>
+  <tr>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">U</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">N</td>
     <td style="text-align: center; border: 1px solid grey; padding: 8px;">Bhattacharyya distance</td>

docs/source/api_reference/detectors/data_drift/batch.md

@@ -47,6 +47,7 @@ The {mod}`frouros.detectors.data_drift.batch` module contains batch data drift d
     :toctree: auto_generated/
     :template: class.md
 
+    AndersonDarlingTest
     ChiSquareTest
     CVMTest
     KSTest

frouros/detectors/data_drift/__init__.py

@@ -1,6 +1,7 @@
 """Data drift detection methods init."""
 
 from .batch import (  # noqa: F401
+    AndersonDarlingTest,
     BhattacharyyaDistance,
     ChiSquareTest,
     CVMTest,
@@ -19,6 +20,7 @@
 from .streaming import IncrementalKSTest, MMD as MMDStreaming  # noqa: N811
 
 __all__ = [
+    "AndersonDarlingTest",
     "BhattacharyyaDistance",
     "ChiSquareTest",
     "CVMTest",

frouros/detectors/data_drift/batch/__init__.py

@@ -11,6 +11,7 @@
     MMD,
 )
 from .statistical_test import (
+    AndersonDarlingTest,
     ChiSquareTest,
     CVMTest,
     KSTest,
@@ -19,6 +20,7 @@
 )
 
 __all__ = [
+    "AndersonDarlingTest",
     "BhattacharyyaDistance",
     "ChiSquareTest",
     "CVMTest",

frouros/detectors/data_drift/batch/statistical_test/anderson_darling.py

@@ -0,0 +1,59 @@
+"""Anderson-Darling test module."""
+
+from typing import Optional, List, Union
+
+import numpy as np  # type: ignore
+from scipy.stats import anderson_ksamp  # type: ignore
+
+from frouros.callbacks.batch.base import BaseCallbackBatch
+from frouros.detectors.data_drift.base import NumericalData, UnivariateData
+from frouros.detectors.data_drift.batch.statistical_test.base import (
+    BaseStatisticalTest,
+    StatisticalResult,
+)
+
+
+class AndersonDarlingTest(BaseStatisticalTest):
+    """Anderson-Darling test [scholz1987k]_ detector.
+
+    :Note:
+     p-values are bounded between 0.001 and 0.25 according to scipy documentation [1]_.
+
+    :References:
+
+    .. [scholz1987k] Scholz, Fritz W., and Michael A. Stephens.
+        "K-sample Anderson–Darling tests."
+        Journal of the American Statistical Association 82.399 (1987): 918-924.
+       [1] https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.anderson_ksamp.html  # noqa: E501 # pylint: disable=line-too-long
+    """
+
+    def __init__(
+        self,
+        callbacks: Optional[Union[BaseCallbackBatch, List[BaseCallbackBatch]]] = None,
+    ) -> None:
+        """Init method.
+
+        :param callbacks: callbacks
+        :type callbacks: Optional[Union[BaseCallbackBatch, List[BaseCallbackBatch]]]
+        """
+        super().__init__(
+            data_type=NumericalData(),
+            statistical_type=UnivariateData(),
+            callbacks=callbacks,
+        )
+
+    def _statistical_test(
+        self, X_ref: np.ndarray, X: np.ndarray, **kwargs  # noqa: N803
+    ) -> StatisticalResult:
+        test = anderson_ksamp(
+            samples=[
+                X_ref,
+                X,
+            ],
+            **kwargs,
+        )
+        test = StatisticalResult(
+            statistic=test.statistic,
+            p_value=test.pvalue,
+        )
+        return test