def test_fit(self):
"""On fit, the attribute model is set with an instance of scipy.stats.truncnorm."""
# Setup
instance = TruncatedGaussian()
# Generate data from a truncated gaussian
loc, scale = 2.0, 1.0
min_value, max_value = -1.0, 4.0
a = (min_value - loc) / scale
b = (max_value - loc) / scale
data = truncnorm.rvs(a=a,
b=b,
loc=loc,
scale=scale,
size=1000000,
random_state=42)
# Run
instance.fit(data)
# Check
assert isinstance(instance.model, rv_frozen)
self.assertAlmostEqual(min_value, instance.min, places=2)
self.assertAlmostEqual(max_value, instance.max, places=2)
self.assertAlmostEqual(loc, instance.mean, places=2)
self.assertAlmostEqual(scale, instance.std, places=2)
def test__extract_constant(self):
distribution = TruncatedGaussian()
distribution._params = {'a': 1, 'b': 1, 'loc': 1, 'scale': 0}
constant = distribution._extract_constant()
assert 1 == constant
def test__fit(self):
distribution = TruncatedGaussian()
data = truncnorm.rvs(size=10000, a=0, b=3, loc=3, scale=1)
distribution._fit(data)
expected = {'loc': 3, 'scale': 1, 'a': 0, 'b': 3}
for key, value in distribution._params.items():
np.testing.assert_allclose(value, expected[key], atol=0.3)
def test_fit(self):
"""On fit, the attribute model is set with an instance of scipy.stats.truncnorm."""
# Setup
instance = TruncatedGaussian()
data = np.array([0, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 4])
# Run
instance.fit(data)
# Check
assert isinstance(instance.model, rv_frozen)
assert instance.min == 0 - EPSILON
assert instance.max == 4 + EPSILON
assert instance.mean == 2
assert instance.std == 1
def test_init(self):
"""On init, there are no errors and attributes are set as expected."""
# Run / Setup
instance = TruncatedGaussian()
# Check
assert instance.model_class == 'truncnorm'
assert instance.unfittable_model is True
assert instance.probability_density == 'pdf'
assert instance.cumulative_distribution == 'cdf'
assert instance.percent_point == 'ppf'
assert instance.sample == 'rvs'
assert instance.model is None
assert instance.fitted is False
assert instance.constant_value is None
def test_from_dict_unfitted(self):
"""from_dict creates a new instance from a dict of params."""
# Setup
parameters = {
'type': 'copulas.univariate.truncnorm.TruncatedGaussian',
'fitted': False,
}
# Run
instance = TruncatedGaussian.from_dict(parameters)
# Check
assert instance.fitted is False
assert instance.constant_value is None
assert instance.model is None
def test_from_dict_constant(self):
"""from_dict creates a new instance from a dict of params."""
# Setup
parameters = {
'type': 'copulas.univariate.truncnorm.TruncatedGaussian',
'fitted': True,
'min': 10,
'max': 10,
'std': 0,
'mean': 10,
}
# Run
instance = TruncatedGaussian.from_dict(parameters)
# Check
assert instance.fitted is True
assert instance.constant_value == 10
def test__fit_params(self):
"""_fit_params returns a dict with the params of the scipy model."""
# Setup
instance = TruncatedGaussian()
instance.min = 0
instance.max = 4
instance.std = 1
instance.mean = 2
expected_result = {
'min': 0,
'max': 4,
'std': 1,
'mean': 2,
}
# Run
result = instance._fit_params()
# Check
assert result == expected_result
def test_from_dict_fitted(self):
"""from_dict creates a new instance from a dict of params."""
# Setup
parameters = {
'type': 'copulas.univariate.truncnorm.TruncatedGaussian',
'fitted': True,
'min': -10,
'max': 10,
'std': 1,
'mean': 1,
}
# Run
instance = TruncatedGaussian.from_dict(parameters)
# Check
assert instance.fitted is True
assert isinstance(instance.model, rv_frozen)
assert instance.constant_value is None
assert instance.min == -10
assert instance.max == 10
assert instance.std == 1
assert instance.mean == 1
def test__fit_constant(self):
distribution = TruncatedGaussian()
distribution._fit_constant(np.array([1, 1, 1, 1]))
assert distribution._params == {'a': 1, 'b': 1, 'loc': 1, 'scale': 0}
def test__is_constant_false(self):
distribution = TruncatedGaussian()
distribution.fit(np.array([1, 2, 3, 4]))
assert not distribution._is_constant()
def test__is_constant_true(self):
distribution = TruncatedGaussian()
distribution.fit(np.array([1, 1, 1, 1]))
assert distribution._is_constant()