def test_midcovariance_shape():
"""
Test that biweight_midcovariance raises error with a 3D array.
"""
d = np.ones(27).reshape(3, 3, 3)
with pytest.raises(ValueError) as e:
biweight_midcovariance(d)
assert 'The input array must be 2D or 1D.' in str(e.value)
def test_midcovariance_shape():
"""
Test that biweight_midcovariance raises error with a 3D array.
"""
d = np.ones(27).reshape(3, 3, 3)
with pytest.raises(ValueError) as e:
biweight_midcovariance(d)
assert 'The input array must be 2D or 1D.' in str(e.value)
def test_midcovariance_M_shape():
"""
Test that biweight_midcovariance raises error when M is not a scalar
or 1D array.
"""
d = [0, 1, 2]
M = [[0, 1], [2, 3]]
with pytest.raises(ValueError) as e:
biweight_midcovariance(d, M=M)
assert 'M must be a scalar or 1D array.' in str(e.value)
def test_midcovariance_M_shape():
"""
Test that biweight_midcovariance raises error when M is not a scalar
or 1D array.
"""
d = [0, 1, 2]
M = [[0, 1], [2, 3]]
with pytest.raises(ValueError) as e:
biweight_midcovariance(d, M=M)
assert 'M must be a scalar or 1D array.' in str(e.value)
def test_biweight_midcovariance_symmetric():
"""
Regression test to ensure that midcovariance matrix is symmetric
when ``modify_sample_size=True`` (see #5972).
"""
rng = np.random.RandomState(1)
d = rng.gamma(2, 2, size=(3, 500))
cov = biweight_midcovariance(d)
assert_array_almost_equal_nulp(cov, cov.T, nulp=5)
cov = biweight_midcovariance(d, modify_sample_size=True)
assert_array_almost_equal_nulp(cov, cov.T, nulp=5)
def test_biweight_midcovariance_symmetric():
"""
Regression test to ensure that midcovariance matrix is symmetric
when ``modify_sample_size=True`` (see #5972).
"""
rng = np.random.RandomState(1)
d = rng.gamma(2, 2, size=(3, 500))
cov = biweight_midcovariance(d)
assert_array_almost_equal_nulp(cov, cov.T, nulp=5)
cov = biweight_midcovariance(d, modify_sample_size=True)
assert_array_almost_equal_nulp(cov, cov.T, nulp=5)
def test_biweight_midcovariance_2d():
d = [[0, 1, 2], [2, 1, 0]]
cov = biweight_midcovariance(d)
val = 0.70121809
assert_allclose(cov, [[val, -val], [-val, val]]) # verified with R
d = [[5, 1, 10], [500, 5, 2]]
cov = biweight_midcovariance(d)
assert_allclose(cov, [[14.54159077, -7.79026256], # verified with R
[-7.79026256, 6.92087252]])
cov = biweight_midcovariance(d, modify_sample_size=True)
assert_allclose(cov, [[14.54159077, -5.19350838],
[-5.19350838, 4.61391501]])
def test_biweight_midcovariance_2d():
d = [[0, 1, 2], [2, 1, 0]]
cov = biweight_midcovariance(d)
val = 0.70121809
assert_allclose(cov, [[val, -val], [-val, val]]) # verified with R
d = [[5, 1, 10], [500, 5, 2]]
cov = biweight_midcovariance(d)
assert_allclose(cov, [[14.54159077, -7.79026256], # verified with R
[-7.79026256, 6.92087252]])
cov = biweight_midcovariance(d, modify_sample_size=True)
assert_allclose(cov, [[14.54159077, -5.19350838],
[-5.19350838, 4.61391501]])
def test_biweight_midcovariance_constant():
data = np.ones((3, 10))
val3 = 5.0
data[1] = [val3, 0.8, val3, -0.8, val3, val3, val3, 1.0, val3, -0.7]
cov = biweight_midcovariance(data)
assert_allclose(cov, np.zeros((3, 3)))
rng = np.random.default_rng(123)
data = rng.random((5, 5))
val3 = 5.0
data[1] = [val3, 0.8, val3, -0.8, val3]
cov = biweight_midcovariance(data)
assert_allclose(cov[1, :], 0.)
assert_allclose(cov[:, 1], 0.)
def test_biweight_midcovariance_midvariance():
"""
Test that biweight_midcovariance diagonal elements agree with
biweight_midvariance.
"""
rng = np.random.RandomState(1)
d = rng.normal(0, 2, size=(100, 3))
cov = biweight_midcovariance(d)
var = [biweight_midvariance(a) for a in d]
assert_allclose(cov.diagonal(), var)
cov2 = biweight_midcovariance(d, modify_sample_size=True)
var2 = [biweight_midvariance(a, modify_sample_size=True) for a in d]
assert_allclose(cov2.diagonal(), var2)
def test_biweight_midcovariance_midvariance():
"""
Test that biweight_midcovariance diagonal elements agree with
biweight_midvariance.
"""
rng = np.random.RandomState(1)
d = rng.normal(0, 2, size=(100, 3))
cov = biweight_midcovariance(d)
var = [biweight_midvariance(a) for a in d]
assert_allclose(cov.diagonal(), var)
cov2 = biweight_midcovariance(d, modify_sample_size=True)
var2 = [biweight_midvariance(a, modify_sample_size=True)
for a in d]
assert_allclose(cov2.diagonal(), var2)
def get_mean_cov(X, robust=True):
"""Get Mean Covariance.
Parameters
----------
X : array-like
robust : bool
Returns
-------
mean_tot : array-like
cov_tot : array-like
"""
from astropy.stats.biweight import (
biweight_midcovariance,
biweight_location,
)
if robust:
# Robust
mean_tot = biweight_location(X, axis=0)
cov_tot = biweight_midcovariance(X.T)
else:
# Classical
mean_tot = np.mean(X, axis=0)
cov_tot = np.cov(X)
return mean_tot, cov_tot
def test_biweight_midcovariance_constant():
data = np.ones((3, 10))
cov = biweight_midcovariance(data)
assert_allclose(cov, np.zeros((3, 3)))
def test_biweight_midcovariance_1d():
d = [0, 1, 2]
cov = biweight_midcovariance(d)
var = biweight_midvariance(d)
assert_allclose(cov, [[var]])
def test_biweight_midcovariance_constant():
data = np.ones((3, 10))
cov = biweight_midcovariance(data)
assert_allclose(cov, np.zeros((3, 3)))
def test_biweight_midcovariance_1d():
d = [0, 1, 2]
cov = biweight_midcovariance(d)
var = biweight_midvariance(d)
assert_allclose(cov, [[var]])
