def test_small_data(self):
# pylint: disable=no-self-use
"""
Tests that all methods return valid data.
Valid data is defined as not all values equal to 0.
"""
# Initialize parameters.
dims = 3
etas = 2
n_samples = 1
beta_params = (50, 2)
# Fixed samples for each method.
cvine_sample = [[[1.0, -0.44642099, 0.02443349],
[-0.44642099, 1.0, -0.72701905],
[0.02443349, -0.72701905, 1.0]]]
ext_onion_sample = [[[1.0, 0.78932386, -0.56877586],
[0.78932386, 1.0, -0.47447423],
[-0.56877586, -0.47447423, 1.0]]]
dvine_sample = [[[1.0, 0.30937182, 0.78298863],
[0.30937182, 1.0, 0.63411242],
[0.78298863, 0.63411242, 1.0]]]
dvine_beta_sample = [[[1.0, 0.9743718, 0.8824407],
[0.9743718, 1.0, 0.86273572],
[0.8824407, 0.86273572, 1.0]]]
cvine_beta_sample = [[[1.0, 0.9282246, 0.95633512],
[0.9282246, 1.0, 0.92777712],
[0.95633512, 0.92777712, 1.0]]]
# Test all methods.
corr_mats = sample_from_cvine(dim=dims, eta=etas, n_samples=n_samples)
np.testing.assert_array_almost_equal(corr_mats, cvine_sample, 5)
corr_mats = sample_from_ext_onion(dim=dims,
eta=etas,
n_samples=n_samples)
np.testing.assert_array_almost_equal(corr_mats, ext_onion_sample, 5)
corr_mats = sample_from_dvine(dim=dims, n_samples=n_samples)
np.testing.assert_array_almost_equal(corr_mats, dvine_sample, 5)
# Test fixed beta parameters for D-vine.
corr_mats = sample_from_dvine(dim=dims,
n_samples=n_samples,
beta_dist_fixed=beta_params)
np.testing.assert_array_almost_equal(corr_mats, dvine_beta_sample, 5)
# Test fixed beta parameters for C-vine.
corr_mats = sample_from_cvine(dim=dims,
eta=etas,
n_samples=n_samples,
beta_dist_fixed=beta_params)
np.testing.assert_array_almost_equal(corr_mats, cvine_beta_sample, 5)
def test_data_returned(self):
"""
Tests that data generated from all methods exists for multiple dimensions and
parameters.
"""
# Test all methods.
for func in [
sample_from_cvine, sample_from_ext_onion, sample_from_dvine
]:
if func == sample_from_dvine:
corr_mat = func(dim=self.dims[0], n_samples=self.n_samples[0])
else:
corr_mat = func(dim=self.dims[0],
eta=self.etas[0],
n_samples=self.n_samples[0])
self.assertTrue(self._data_exists(corr_mat),
msg="{} failed".format(func.__name__))
# Test fixed beta parameters for D-vine.
corr_mat = sample_from_dvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_exists(corr_mat))
# Test fixed beta parameters for C-vine.
corr_mat = sample_from_cvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_exists(corr_mat))
def test_matrix_is_symmetrical(self):
"""
Tests that the matrices from all methods generated are symmetrical.
"""
# Test all methods.
for func in [
sample_from_cvine, sample_from_ext_onion, sample_from_dvine
]:
for i in range(self.n_generated):
if func == sample_from_dvine:
corr_mats = func(dim=self.dims[i],
n_samples=self.n_samples[i])
else:
corr_mats = func(dim=self.dims[i],
eta=self.etas[i],
n_samples=self.n_samples[i])
self.assertTrue(
self._data_is_symmetric(corr_mats, self.dims[i]),
msg="{} failed".format(func.__name__),
)
# Test fixed beta parameters for D-vine.
corr_mats = sample_from_dvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_is_symmetric(corr_mats, self.dims[0]))
# Test fixed beta parameters for C-vine.
corr_mats = sample_from_cvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_is_symmetric(corr_mats, self.dims[0]))
def test_data_generated_is_correctly_bounded(self):
"""
Tests that the data generated from all methods is bounded to the range
-1 <= data <= 1 which corresponds to any correlation matrix bounds,
and that the diagonal is always equal to 1s.
"""
# Test all methods.
for func in [
sample_from_cvine, sample_from_ext_onion, sample_from_dvine
]:
for i in range(self.n_generated):
if func == sample_from_dvine:
corr_mats = func(dim=self.dims[i],
n_samples=self.n_samples[i])
else:
corr_mats = func(dim=self.dims[i],
eta=self.etas[i],
n_samples=self.n_samples[i])
self.assertTrue(
self._data_correctly_bounded(corr_mats, self.dims[i]),
msg="{} failed".format(func.__name__),
)
# Test fixed beta parameters for D-vine.
corr_mats = sample_from_dvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_correctly_bounded(corr_mats, self.dims[0]))
# Test fixed beta parameters for C-vine.
corr_mats = sample_from_cvine(dim=self.dims[0],
eta=self.etas[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(self._data_correctly_bounded(corr_mats, self.dims[0]))
def test_correct_shapes(self):
"""
Tests that data generated from all methods exists for multiple dimensions and
parameters.
"""
# Test all methods.
for func in [
sample_from_cvine, sample_from_ext_onion, sample_from_dvine
]:
for i in range(self.n_generated):
if func == sample_from_dvine:
corr_mat = func(dim=self.dims[i],
n_samples=self.n_samples[i])
else:
corr_mat = func(dim=self.dims[i],
eta=self.etas[i],
n_samples=self.n_samples[i])
self.assertTrue(
corr_mat.shape == (self.n_samples[i], self.dims[i],
self.dims[i]),
msg="{} failed".format(func.__name__),
)
# Test fixed beta parameters for D-vine.
corr_mat = sample_from_dvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(corr_mat.shape == (self.n_samples[0], self.dims[0],
self.dims[0]))
# Test fixed beta parameters for C-vine.
corr_mat = sample_from_cvine(dim=self.dims[0],
n_samples=self.n_samples[0],
beta_dist_fixed=self.beta_params)
self.assertTrue(corr_mat.shape == (self.n_samples[0], self.dims[0],
self.dims[0]))