Source code for gammapy.stats.utils

# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Utility functions."""
from __future__ import absolute_import, division, print_function, unicode_literals
import numpy as np

__all__ = ['cov_to_corr']


def cov_to_corr(covariance):
    """
    Compute correlation matrix from covariance matrix.

    The correlation matrix :math:`c` is related to the covariance matrix :math:`\\sigma` by:

    .. math::

        c_{ij} = \\frac{\\sigma_{ij}}{\\sqrt{\\sigma_{ii} \sigma_{jj}}}

    Parameters
    ----------
    covariance : `numpy.array`
        Covariance matrix.

    Returns
    -------
    correlation : `numpy.array`
        Correlation matrix.
    """
    diagonal = np.sqrt(covariance.diagonal())
    return (covariance.T / diagonal).T / diagonal