model-diagnostics 1.4.3__tar.gz → 1.5.0__tar.gz

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/.gitignore

@@ -120,7 +120,7 @@ venv.bak/
 # Rope project settings
 .ropeproject
 
-# mkdocs documentation
+# properdocs documentation
 /site
 
 # mypy

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: model-diagnostics
-Version: 1.4.3
+Version: 1.5.0
 Summary: Tools for diagnostics and assessment of (machine learning) models
 Project-URL: Homepage, https://lorentzenchr.github.io/model-diagnostics/
 Project-URL: Source, https://github.com/lorentzenchr/model-diagnostics
@@ -48,7 +48,7 @@ Description-Content-Type: text/markdown
 
 | | |
 | --- | --- |
-| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics)
+| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics) [![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/lorentzenchr/model-diagnostics/badge)](https://scorecard.dev/viewer/?uri=github.com/lorentzenchr/model-diagnostics)
 | Docs | [![Docs](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml)
 | Package | [![PyPI - Version](https://img.shields.io/pypi/v/model-diagnostics.svg?logo=pypi&label=PyPI&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Downloads](https://img.shields.io/pypi/dm/model-diagnostics.svg?color=blue&label=Downloads&logo=pypi&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/model-diagnostics.svg?logo=python&label=Python&logoColor=gold)](https://pypi.org/project/model-diagnostics/) |
 | Meta | [![Hatch project](https://img.shields.io/badge/%F0%9F%A5%9A-Hatch-4051b5.svg)](https://github.com/pypa/hatch) [![linting - Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff) [![types - Mypy](https://img.shields.io/badge/types-Mypy-blue.svg)](https://github.com/python/mypy) [![License - MIT](https://img.shields.io/badge/license-MIT-9400d3.svg)](https://spdx.org/licenses/)
@@ -58,14 +58,14 @@ Description-Content-Type: text/markdown
 Highlights:
 
 - All common point predictions covered: mean, median, quantiles, expectiles.
-- Assess model calibration with [identification functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.identification_function) (generalized residuals), [compute_bias](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.compute_bias) and [compute_marginal](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.compute_marginal).
+- Assess model calibration with [identification functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.identification_function) (generalized residuals), [compute_bias](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.compute_bias) and [compute_marginal](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.compute_marginal).
 - Assess calibration and bias graphically
-    - [reliability diagrams](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_reliability_diagram) for auto-calibration
-    - [bias plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_bias) for conditional calibration
-    - [marginal plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_marginal) for average `y_obs`, `y_pred` and partial dependence for one feature
+    - [reliability diagrams](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_reliability_diagram) for auto-calibration
+    - [bias plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_bias) for conditional calibration
+    - [marginal plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_marginal) for average `y_obs`, `y_pred` and partial dependence for one feature
 - Assess the predictive performance of models
-    - strictly consistent, homogeneous [scoring functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/scoring/)
-    - [score decomposition](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/scoring/#model_diagnostics.scoring.scoring.decompose) into miscalibration, discrimination and uncertainty
+    - strictly consistent, homogeneous [scoring functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/)
+    - [score decomposition](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/#model_diagnostics.scoring.decompose) into miscalibration, discrimination and uncertainty
 - Choose your plot backend, either [matplotlib](https://matplotlib.org) or [plotly](https://plotly.com/python/), e.g., via [set_config](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/#model_diagnostics.set_config).
 
 :rocket: To our knowledge, this is the first python package to offer reliability diagrams for quantiles and expectiles and a score decomposition, both made available by an internal implementation of isotonic quantile/expectile regression. :rocket:

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/README.md

@@ -2,7 +2,7 @@
 
 | | |
 | --- | --- |
-| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics)
+| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics) [![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/lorentzenchr/model-diagnostics/badge)](https://scorecard.dev/viewer/?uri=github.com/lorentzenchr/model-diagnostics)
 | Docs | [![Docs](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml)
 | Package | [![PyPI - Version](https://img.shields.io/pypi/v/model-diagnostics.svg?logo=pypi&label=PyPI&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Downloads](https://img.shields.io/pypi/dm/model-diagnostics.svg?color=blue&label=Downloads&logo=pypi&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/model-diagnostics.svg?logo=python&label=Python&logoColor=gold)](https://pypi.org/project/model-diagnostics/) |
 | Meta | [![Hatch project](https://img.shields.io/badge/%F0%9F%A5%9A-Hatch-4051b5.svg)](https://github.com/pypa/hatch) [![linting - Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff) [![types - Mypy](https://img.shields.io/badge/types-Mypy-blue.svg)](https://github.com/python/mypy) [![License - MIT](https://img.shields.io/badge/license-MIT-9400d3.svg)](https://spdx.org/licenses/)
@@ -12,14 +12,14 @@
 Highlights:
 
 - All common point predictions covered: mean, median, quantiles, expectiles.
-- Assess model calibration with [identification functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.identification_function) (generalized residuals), [compute_bias](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.compute_bias) and [compute_marginal](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/identification/#model_diagnostics.calibration.identification.compute_marginal).
+- Assess model calibration with [identification functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.identification_function) (generalized residuals), [compute_bias](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.compute_bias) and [compute_marginal](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.compute_marginal).
 - Assess calibration and bias graphically
-    - [reliability diagrams](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_reliability_diagram) for auto-calibration
-    - [bias plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_bias) for conditional calibration
-    - [marginal plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/plots/#model_diagnostics.calibration.plots.plot_marginal) for average `y_obs`, `y_pred` and partial dependence for one feature
+    - [reliability diagrams](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_reliability_diagram) for auto-calibration
+    - [bias plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_bias) for conditional calibration
+    - [marginal plots](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/calibration/#model_diagnostics.calibration.plot_marginal) for average `y_obs`, `y_pred` and partial dependence for one feature
 - Assess the predictive performance of models
-    - strictly consistent, homogeneous [scoring functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/scoring/)
-    - [score decomposition](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/scoring/#model_diagnostics.scoring.scoring.decompose) into miscalibration, discrimination and uncertainty
+    - strictly consistent, homogeneous [scoring functions](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/)
+    - [score decomposition](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/scoring/#model_diagnostics.scoring.decompose) into miscalibration, discrimination and uncertainty
 - Choose your plot backend, either [matplotlib](https://matplotlib.org) or [plotly](https://plotly.com/python/), e.g., via [set_config](https://lorentzenchr.github.io/model-diagnostics/reference/model_diagnostics/#model_diagnostics.set_config).
 
 :rocket: To our knowledge, this is the first python package to offer reliability diagrams for quantiles and expectiles and a score decomposition, both made available by an internal implementation of isotonic quantile/expectile regression. :rocket:

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/docs/gen_ref_pages.py

@@ -17,7 +17,7 @@ log_msg = ["\n\tAPI tReference:"]
 for path in sorted(
     set(src.rglob("*.py"))
     - set(src.rglob("*tests/*.py"))
-    # - set(src.rglob("*/_[!_]*.py"))  # No single underscore file
+    - set(src.rglob("*/_[!_]*.py"))  # No single underscore file
     - set(src.rglob("*/_[!_]*/*.py"))  # No single underscore module
     - set(src.rglob("__about__.py"))
 ):

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/docs/index.md

@@ -2,7 +2,7 @@
 
 | | |
 | --- | --- |
-| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics)
+| CI/CD |[![CI - Test](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/test.yml) [![Coverage](https://codecov.io/github/lorentzenchr/model-diagnostics/coverage.svg?branch=main)](https://codecov.io/gh/lorentzenchr/model-diagnostics) [![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/lorentzenchr/model-diagnostics/badge)](https://scorecard.dev/viewer/?uri=github.com/lorentzenchr/model-diagnostics)
 | Docs | [![Docs](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml/badge.svg)](https://github.com/lorentzenchr/model-diagnostics/actions/workflows/docs.yml)
 | Package | [![PyPI - Version](https://img.shields.io/pypi/v/model-diagnostics.svg?logo=pypi&label=PyPI&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Downloads](https://img.shields.io/pypi/dm/model-diagnostics.svg?color=blue&label=Downloads&logo=pypi&logoColor=gold)](https://pypi.org/project/model-diagnostics/) [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/model-diagnostics.svg?logo=python&label=Python&logoColor=gold)](https://pypi.org/project/model-diagnostics/) |
 | Meta | [![Hatch project](https://img.shields.io/badge/%F0%9F%A5%9A-Hatch-4051b5.svg)](https://github.com/pypa/hatch) [![linting - Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff) [![types - Mypy](https://img.shields.io/badge/types-Mypy-blue.svg)](https://github.com/python/mypy) [![License - MIT](https://img.shields.io/badge/license-MIT-9400d3.svg)](https://spdx.org/licenses/)
@@ -12,13 +12,13 @@
 Highlights:
 
 - All common point predictions covered: mean, median, quantiles, expectiles.
-- Assess model calibration with [identification functions][model_diagnostics.calibration.identification.identification_function] (generalized residuals), [compute_bias][model_diagnostics.calibration.identification.compute_bias] and [compute_marginal][model_diagnostics.calibration.identification.compute_marginal].
+- Assess model calibration with [identification functions][model_diagnostics.calibration.identification_function] (generalized residuals), [compute_bias][model_diagnostics.calibration.compute_bias] and [compute_marginal][model_diagnostics.calibration.compute_marginal].
 - Assess calibration and bias graphically
-    - [reliability diagrams][model_diagnostics.calibration.plots.plot_reliability_diagram] for auto-calibration
-    - [bias plots][model_diagnostics.calibration.plots.plot_bias] for conditional calibration
-    - [marginal plots][model_diagnostics.calibration.plots.plot_marginal] for average `y_obs`, `y_pred` and partial dependence for one feature
+    - [reliability diagrams][model_diagnostics.calibration.plot_reliability_diagram] for auto-calibration
+    - [bias plots][model_diagnostics.calibration.plot_bias] for conditional calibration
+    - [marginal plots][model_diagnostics.calibration.plot_marginal] for average `y_obs`, `y_pred` and partial dependence for one feature
 - Assess the predictive performance of models
-    - strictly consistent, homogeneous [scoring functions][model_diagnostics.scoring.scoring]
+    - strictly consistent, homogeneous [scoring functions][model_diagnostics.scoring]
     - [score decomposition][model_diagnostics.scoring.decompose] into miscalibration, discrimination and uncertainty
 - Choose your plot backend, either [matplotlib](https://matplotlib.org) or [plotly](https://plotly.com/python/), e.g., via [set_config][model_diagnostics.set_config].
 

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/pyproject.toml

@@ -166,7 +166,8 @@ no-cov = "cov --no-cov"
 
 [tool.hatch.envs.docs]
 dependencies = [
-  "mkdocs-material~=9.5.31",
+  "properdocs>=1.6.7",  # https://github.com/ProperDocs/properdocs/discussions/33
+  "mkdocs-material~=9.6.7",
   "mkdocstrings[python]>=0.25.1",
   "mkdocs-gen-files>=0.5.0",
   "mkdocs-literate-nav>=0.6.1",
@@ -176,9 +177,9 @@ dependencies = [
 ]
 
 [tool.hatch.envs.docs.scripts]
-build = "mkdocs build --clean --strict {args}"
-serve = "mkdocs serve --dev-addr localhost:8000 {args}"
-ci-build = "mkdocs gh-deploy --force {args}"
+build = "properdocs build --clean --strict {args}"
+serve = "properdocs serve --dev-addr localhost:8000 {args}"
+ci-build = "properdocs gh-deploy --force {args}"
 
 [tool.hatch.envs.jupyter]
 extra-dependencies = [
@@ -297,4 +298,5 @@ doctest_optionflags = [
 filterwarnings = [
   # fail on any warnings that are not explicitly matched below
   "error",
+  "ignore:Support for the dataframe interchange protocol:DeprecationWarning",
 ]

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/src/model_diagnostics/__about__.py

@@ -14,4 +14,4 @@
 # Dev branch marker is: 'X.Y.dev' or 'X.Y.devN' where N is an integer.
 # 'X.Y.dev0' is the canonical version of 'X.Y.dev'
 #
-__version__ = "1.4.3"
+__version__ = "1.5.0"

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/src/model_diagnostics/_utils/array.py

@@ -1,13 +1,10 @@
 import copy
 import sys
-import warnings
-from importlib.metadata import version
 from typing import Optional, Union
 
 import numpy as np
 import numpy.typing as npt
 import polars as pl
-from packaging.version import Version, parse
 
 AL_or_polars = Union[npt.ArrayLike, pl.Series]
 
@@ -261,46 +258,32 @@ def safe_assign_column(x, values, column_index):
             #   Setting an item of incompatible dtype is deprecated and will raise in a
             #   future error of pandas.
             # Also, assigning with a different index makes troubles.
+            # Also, we avoid to have pyarrow installed, which newer versions of polars
+            # need to series.to_pandas().
             pd = sys.modules["pandas"]
             dtype = x.dtypes.iloc[column_index]
-            if isinstance(values, pl.Series) and isinstance(
-                values.dtype, pl.Categorical
-            ):
-                # FIXME: pyarrow not installed
-                pd_values = pd.Series(
-                    data=values.cast(pl.Utf8).to_numpy(),
-                    dtype=dtype,
-                )
+            if isinstance(values, pl.Series):
+                if isinstance(values.dtype, (pl.Categorical, pl.String)):
+                    pd_values = pd.Series(
+                        data=values.cast(pl.String).to_numpy(),
+                        dtype=dtype,
+                    )
+                elif values.dtype.is_integer() or values.dtype.is_float():
+                    pd_values = pd.Series(
+                        data=values.to_numpy(),
+                        dtype=dtype,
+                    )
+                else:
+                    pd_values = pd.Series(
+                        data=values.to_pandas(),
+                        dtype=dtype,
+                    )
             else:
                 pd_values = pd.Series(
-                    data=values.to_pandas()
-                    if isinstance(values, pl.Series)
-                    else values,
+                    data=values,
                     dtype=dtype,
                 )
-            if parse(version("pandas")) < Version("2.0.0"):
-                # FIXME: pandas >= 2.0 (<2.0 means 1.5.*)
-                with warnings.catch_warnings():
-                    msg = (
-                        r"In a future version, `df.iloc\[:, i\] = newvals` will "
-                        r"attempt to set the values inplace instead of always "
-                        r"setting a new array. To retain the old behavior, use either "
-                        r"`df\[df.columns\[i\]\] = newvals` or, if columns are "
-                        r"non-unique, `df.isetitem\(i, newvals\)`"
-                    )
-                    warnings.filterwarnings(
-                        "ignore", category=DeprecationWarning, message=msg
-                    )
-                    if not x.index.is_unique:
-                        # Pandas might error with:
-                        #   cannot reindex on an axis with duplicate labels
-                        # Try reindexing ourselves.
-                        x = x.reset_index()
-                    if not pd_values.index.is_unique:
-                        pd_values = pd_values.reset_index()
-                    x.iloc[:, column_index] = pd_values
-            else:
-                x.iloc[:, column_index] = pd_values
+            x.iloc[:, column_index] = pd_values
         except Exception as e:
             # FIXME: pyarrow version XXX
             # Older pyarrow versions of AttributeError do not have a 'add_note' method.
@@ -352,3 +335,92 @@ def safe_index_rows(x, indices):
     else:
         # numpy, polars
         return x[indices]
+
+
+def safe_copy(x) -> npt.ArrayLike:
+    """Create a safe copy of input data in various formats.
+
+    'Safe' means that it is guaranteed that the original data will not be modified
+    by modifications of the 'safe' copy. But only copy if really required.
+
+    Parameters
+    ----------
+    x : array-like
+        The input data to be copied, which can be numpy array, pandas DataFrame,
+        polars DataFrame, PyArrow Table, or other object types.
+
+    Returns
+    -------
+    copied_x : array-like
+        A copy of the input data in the same format.
+    """
+    if is_pyarrow_table(x) or isinstance(x, pl.DataFrame):
+        # Copy on Write
+        pass
+    elif hasattr(x, "copy"):
+        # list, numpy, pandas, scipy sparse, ...
+        x = x.copy()
+    elif not isinstance(x, tuple):
+        x = copy.deepcopy(x)
+    return x
+
+
+def get_column_names(x) -> list:
+    """Extract column names from different data containers.
+
+    This function handles different data container formats and returns
+    their column names or indices if names are not available.
+
+    Parameters
+    ----------
+    x : array-like
+        The input data which can be a numpy array, pandas DataFrame,
+        polars DataFrame, PyArrow Table, or other similar data container.
+
+    Returns
+    -------
+    list
+        A list of column names if available, or column indices (integers)
+        for array-like objects without named columns.
+    """
+    if is_pyarrow_table(x):
+        colnames = x.column_names
+    elif is_pandas_df(x):
+        colnames = x.columns.to_list()
+    elif hasattr(x, "columns") and isinstance(x.columns, list):
+        # polars
+        colnames = x.columns
+    else:
+        # e.g. numpy
+        colnames = [str(i) for i in range(x.shape[1])]
+    return colnames
+
+
+def to_pl_series(x, name=None) -> pl.Series:
+    """Convert x to a polars Series."""
+    try:
+        feature_out = pl.Series(name=name, values=x)
+    except ImportError:
+        # FIXME: pyarrow not installed
+        # For non numpy-backed columns, pyarrow is needed. Here we handle the case
+        # where pyarrow is not installed and such a pandas extention array is
+        # passed, e.g. with CategoricalDtype.
+        if is_pandas_series(x):
+            pandas = sys.modules["pandas"]
+            is_pandas_categorical = isinstance(
+                x.dtype,  # type: ignore
+                pandas.CategoricalDtype,
+            )
+            feature_out = pl.from_dataframe(
+                x.to_frame(name=name)  # type: ignore
+            )[:, 0]
+            if name is None:
+                feature_out = feature_out.alias("")
+            if is_pandas_categorical and isinstance(feature_out.dtype, pl.Enum):
+                # Pandas categoricals usually get mapped to polars categoricals.
+                # But this code path gives pl.Enum.
+                feature_out = feature_out.cast(pl.Categorical)
+        else:
+            raise  # re-raises the ImportError
+
+    return feature_out

{model_diagnostics-1.4.3 → model_diagnostics-1.5.0}/src/model_diagnostics/_utils/binning.py

@@ -1,4 +1,4 @@
-import sys
+from math import ceil
 from typing import Optional, Union
 
 import numpy as np
@@ -7,8 +7,8 @@ import polars as pl
 
 from model_diagnostics._utils.array import (
     array_name,
-    is_pandas_series,
     length_of_first_dimension,
+    to_pl_series,
 )
 
 
@@ -118,34 +118,15 @@ def bin_feature(
     feature_name = array_name(feature, default=default)
     # The following statement, i.e. possibly the creation of a pl.Categorical,
     # MUST be under the StringCache context manager!
-    try:
-        feature_out = pl.Series(name=feature_name, values=feature)
-    except ImportError:
-        # FIXME: pyarrow not installed
-        # For non numpy-backed columns, pyarrow is needed. Here we handle the case
-        # where pyarrow is not installed and such a pandas extention array is
-        # passed, e.g. with CategoricalDtype.
-        if is_pandas_series(feature):
-            pandas = sys.modules["pandas"]
-            is_pandas_categorical = isinstance(
-                feature.dtype,  # type: ignore
-                pandas.CategoricalDtype,
-            )
-            feature_out = pl.from_dataframe(
-                feature.to_frame(name=feature_name)  # type: ignore
-            )[:, 0]
-            if is_pandas_categorical and isinstance(feature_out.dtype, pl.Enum):
-                # Pandas categoricals usually get mapped to polars categoricals.
-                # But this code path gives pl.Enum.
-                feature_out = feature_out.cast(pl.Categorical)
-        else:
-            raise  # re-raises the ImportError
+    feature_out = to_pl_series(feature, name=feature_name)
+
     if length_of_first_dimension(feature_out) != n_obs:
         msg = (
             f"The feature array {feature_name} does not have length {n_obs} of its"
             " first dimension."
         )
         raise ValueError(msg)
+
     if feature_out.dtype == pl.Categorical:
         is_categorical = True
     elif feature_out.dtype == pl.Enum:
@@ -299,3 +280,34 @@ def bin_feature(
             .collect()
         )
     return feature_out, n_bins_ef + feature_out.has_nulls(), f_binned
+
+
+def compute_grid(x: pl.Series, n: int) -> pl.Series:
+    """Compute a grid with n equal distributed points.
+
+    Best call this function inside a `with pl.StringCache()` context manager.
+    """
+    if x.dtype.is_numeric() or x.dtype.is_temporal():
+        f_min, f_max = x.min(), x.max()
+        if f_min == f_max:
+            return pl.Series([f_min], dtype=x.dtype)
+        elif x.dtype.is_float():
+            # pl.linear_space was added in version 1.21.0, we have minimum 1.1.0.
+            grid = pl.Series(np.linspace(f_min, f_max, n), dtype=x.dtype)
+        elif x.dtype.is_integer():
+            step = max(1, ceil((f_max - f_min + 1) / n))
+            grid = pl.int_range(f_min, f_max + 1, step, dtype=x.dtype, eager=True)
+        elif isinstance(x.dtype, (pl.Date, pl.Datetime, pl.Time)):
+            d_range = {
+                pl.Date: pl.date_range,
+                pl.Datetime: pl.datetime_range,
+                pl.Time: pl.time_range,
+            }
+            n_res = int((f_max - f_min) / f_min.resolution) // max(1, (n - 1))
+            grid = d_range[x.dtype](f_min, f_max, n_res * f_min.resolution, eager=True)
+            # TODO: pl.Duration
+    else:
+        # String, Categorical, Enum or complex types such as List, Union
+        grid = x.value_counts(sort=True).head(n)[:, 0]
+
+    return grid

model_diagnostics-1.5.0/src/model_diagnostics/_utils/identification.py

@@ -0,0 +1,108 @@
+import numpy as np
+import numpy.typing as npt
+
+from model_diagnostics._utils.array import validate_2_arrays
+
+
+def identification_function(
+    y_obs: npt.ArrayLike,
+    y_pred: npt.ArrayLike,
+    *,
+    functional: str = "mean",
+    level: float = 0.5,
+) -> np.ndarray:
+    r"""Canonical identification function.
+
+    Identification functions act as generalised residuals. See [Notes](#notes) for
+    further details.
+
+    Parameters
+    ----------
+    y_obs : array-like of shape (n_obs)
+        Observed values of the response variable.
+        For binary classification, y_obs is expected to be in the interval [0, 1].
+    y_pred : array-like of shape (n_obs)
+        Predicted values of the `functional`, e.g. the conditional expectation of
+        the response, `E(Y|X)`.
+    functional : str
+        The functional that is induced by the identification function `V`. Options are:
+
+        - `"mean"`. Argument `level` is neglected.
+        - `"median"`. Argument `level` is neglected.
+        - `"expectile"`
+        - `"quantile"`
+
+    level : float
+        The level of the expectile of quantile. (Often called \(\alpha\).)
+        It must be `0 < level < 1`.
+        `level=0.5` and `functional="expectile"` gives the mean.
+        `level=0.5` and `functional="quantile"` gives the median.
+
+    Returns
+    -------
+    V : ndarray of shape (n_obs)
+        Values of the identification function.
+
+    Notes
+    -----
+    [](){#notes}
+
+    The function \(V(y, z)\) for observation \(y=y_{pred}\) and prediction
+    \(z=y_{pred}\) is a strict identification function for the functional \(T\), or
+    induces the functional \(T\) as:
+
+    \[
+    \mathbb{E}[V(Y, z)] = 0\quad \Leftrightarrow\quad z\in T(F) \quad \forall
+    \text{ distributions } F
+    \in \mathcal{F}
+    \]
+
+    for some class of distributions \(\mathcal{F}\). Implemented examples of the
+    functional \(T\) are mean, median, expectiles and quantiles.
+
+    | functional | strict identification function \(V(y, z)\)           |
+    | ---------- | ---------------------------------------------------- |
+    | mean       | \(z - y\)                                            |
+    | median     | \(\mathbf{1}\{z \ge y\} - \frac{1}{2}\)              |
+    | expectile  | \(2 \mid\mathbf{1}\{z \ge y\} - \alpha\mid (z - y)\) |
+    | quantile   | \(\mathbf{1}\{z \ge y\} - \alpha\)                   |
+
+    For `level` \(\alpha\).
+
+    References
+    ----------
+    `[Gneiting2011]`
+
+    :   T. Gneiting.
+        "Making and Evaluating Point Forecasts". (2011)
+        [doi:10.1198/jasa.2011.r10138](https://doi.org/10.1198/jasa.2011.r10138)
+        [arxiv:0912.0902](https://arxiv.org/abs/0912.0902)
+
+    Examples
+    --------
+    >>> identification_function(y_obs=[0, 0, 1, 1], y_pred=[-1, 1, 1 , 2])
+    array([-1,  1,  0,  1])
+    """
+    y_o: np.ndarray
+    y_p: np.ndarray
+    y_o, y_p = validate_2_arrays(y_obs, y_pred)
+
+    if functional in ("expectile", "quantile") and (level <= 0 or level >= 1):
+        msg = f"Argument level must fulfil 0 < level < 1, got {level}."
+        raise ValueError(msg)
+
+    if functional == "mean":
+        return y_p - y_o
+    elif functional == "median":
+        return np.greater_equal(y_p, y_o) - 0.5
+    elif functional == "expectile":
+        return 2 * np.abs(np.greater_equal(y_p, y_o) - level) * (y_p - y_o)
+    elif functional == "quantile":
+        return np.greater_equal(y_p, y_o) - level
+    else:
+        allowed_functionals = ("mean", "median", "expectile", "quantile")
+        msg = (
+            f"Argument functional must be one of {allowed_functionals}, got "
+            f"{functional}."
+        )
+        raise ValueError(msg)