ccc.py

# Copyright (C) 2024 Aaron Keesing
#
# Permission is hereby granted, free of charge, to any person obtaining
# a copy of this software and associated documentation files (the
# “Software”), to deal in the Software without restriction, including
# without limitation the rights to use, copy, modify, merge, publish,
# distribute, sublicense, and/or sell copies of the Software, and to
# permit persons to whom the Software is furnished to do so, subject to
# the following conditions:
#
# The above copyright notice and this permission notice shall be
# included in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
# IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
# CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

"""Concordance correlation coefficient"""

import datasets
import evaluate
import numpy as np

_CITATION = """\
@article{linConcordanceCorrelationCoefficient1989,
  title = {A {{Concordance Correlation Coefficient}} to {{Evaluate Reproducibility}}},
  author = {Lin, Lawrence I-Kuei},
  year = {1989},
  journal = {Biometrics},
  volume = {45},
  number = {1},
  pages = {255--268},
  publisher = {{International Biometric Society}},
  issn = {0006-341X},
  url = {https://www.jstor.org/stable/2532051},
  doi = {10.2307/2532051}
}

"""

_DESCRIPTION = """\
A metric to measure the degree of agreement between continuous-values evaluations from two raters.
"""


# TODO: Add description of the arguments of the module here
_KWARGS_DESCRIPTION = """
Calculates the CCC between predictions and references
Args:
    predictions: list of predictions to score. Each prediction
        should be a floating point value.
    references: list of references, one for each prediction. Each
        reference should be a floating point value.
Returns:
    ccc: the concordance correlation coefficient, -1 <= ccc <= 1
Examples:
    >>> ccc_metric = evaluate.load("agkphysics/ccc")
    >>> results = ccc_metric.compute(references=[0.2, 0.1], predictions=[0.1, 0.2])
    >>> print(results)
    {'ccc': -1.0}

    >>> results = ccc_metric.compute(references=[0.1, 0.2], predictions=[0.1, 0.2])
    >>> print(results)
    {'ccc': 1.0}

    >>> results = ccc_metric.compute(references=[0.1, 0.3], predictions=[0.1, 0.2])
    >>> print(results)
    {'ccc': 0.666666641831399}
"""


@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
class CCC(evaluate.Metric):
    """Computes the CCC, concordance correlation coefficient."""

    def _info(self):
        return evaluate.MetricInfo(
            module_type="metric",
            description=_DESCRIPTION,
            citation=_CITATION,
            inputs_description=_KWARGS_DESCRIPTION,
            features=datasets.Features(
                {
                    "predictions": datasets.Value("float32"),
                    "references": datasets.Value("float32"),
                }
            ),
            homepage="https://en.wikipedia.org/wiki/Concordance_correlation_coefficient",
            reference_urls=[
                "https://www.jstor.org/stable/2532051",
                "https://en.wikipedia.org/wiki/Concordance_correlation_coefficient",
            ],
        )

    def _compute(self, predictions, references):
        """Returns the CCC score"""
        sxy = np.cov(predictions, references, ddof=0)[0, 1]
        sxx = np.var(predictions, ddof=0)
        syy = np.var(references, ddof=0)
        mu_x = np.mean(predictions)
        mu_y = np.mean(references)
        ccc = 2 * sxy / (sxx + syy + (mu_x - mu_y) ** 2)
        return {"ccc": ccc}