def test_dataset_init(variables):
dataset = Dataset(*variables, X_names=["bork"])
n = len(variables[0])
assert dataset.X.shape == (n, 2)
assert dataset.X_names == ["intercept", "bork"]
dataset = Dataset(*variables, X_names=["bork"], add_intercept=False)
assert dataset.X.shape == (n, 1)
assert dataset.X_names == ["bork"]
def test_dataset_init(variables):
"""Test Dataset creation from numpy arrays."""
dataset = Dataset(*variables, X_names=["bork"])
n = len(variables[0])
assert dataset.X.shape == (n, 2)
assert dataset.X_names == ["intercept", "bork"]
dataset = Dataset(*variables, X_names=["bork"], add_intercept=False)
assert dataset.X.shape == (n, 1)
assert dataset.X_names == ["bork"]
def test_combination_test_from_dataset(Cls, data, mode, expected):
"""Test CombinationTest Estimators with PyMARE Datasets."""
dset = Dataset(y=data)
est = Cls(mode).fit_dataset(dset)
results = est.summary()
z = ss.norm.isf(results.p)
assert np.allclose(z, expected, atol=1e-5)
def dataset_2d(variables):
y, v, X = variables
y = np.repeat(y, 3, axis=1)
y[:, 1] = np.random.randint(-10, 10, size=len(y))
v = np.repeat(v, 3, axis=1)
v[:, 1] = np.random.randint(2, 10, size=len(v))
return Dataset(y, v, X)
def dataset_n():
y = np.array(
[[-3.0, -0.5, 0.0, -5.01, 0.35, -2.0, -6.0, -4.0, -4.3, -0.1, -1.0]]).T
n = (np.array([[
16, 16, 20.548, 32.597, 14.0, 11.118, 4.444, 12.414, 26.963, 130.556,
126.76
]]).T / 2)
return Dataset(y, n=n)
def dataset_n():
y = np.array([[-3., -0.5, 0., -5.01, 0.35, -2., -6., -4., -4.3, -0.1,
-1.]]).T
n = np.array([[
16, 16, 20.548, 32.597, 14., 11.118, 4.444, 12.414, 26.963, 130.556,
126.76
]]).T / 2
return Dataset(y, sample_sizes=n)
def dataset_2d(variables):
"""Build a larger Dataset with 2D data."""
y, v, X = variables
y = np.repeat(y, 3, axis=1)
y[:, 1] = np.random.randint(-10, 10, size=len(y))
v = np.repeat(v, 3, axis=1)
v[:, 1] = np.random.randint(2, 10, size=len(v))
return Dataset(y, v, X)
def to_dataset(self, measure='g', **kwargs):
y = self.get(measure)
v = self.get('v_{}'.format(measure), error=False)
try:
n = self.get('n')
except:
n = None
return Dataset(y=y, v=v, n=n, **kwargs)
def to_dataset(self, measure, **kwargs):
measure = measure.lower()
y = self.get(measure)
v = self.get("v_{}".format(measure), error=False)
try:
n = self.get("n")
except:
n = None
return Dataset(y=y, v=v, n=n, **kwargs)
def dataset_n():
"""Build a Dataset with sample sizes, but no variances."""
y = np.array(
[[-3.0, -0.5, 0.0, -5.01, 0.35, -2.0, -6.0, -4.0, -4.3, -0.1, -1.0]]).T
n = (np.array([[
16, 16, 20.548, 32.597, 14.0, 11.118, 4.444, 12.414, 26.963, 130.556,
126.76
]]).T / 2)
return Dataset(y, n=n)
def test_exact_perm_test_1d_no_mods():
dataset = Dataset([1, 1, 2, 1.3], [1.5, 1, 2, 4])
results = DerSimonianLaird().fit(dataset).summary()
pmr = permutation_test(results, 867)
assert pmr.n_perm == 16
assert pmr.exact
assert isinstance(pmr.results, MetaRegressionResults)
assert pmr.fe_p.shape == (1, 1)
assert pmr.tau2_p.shape == (1, )
def to_dataset(self, measure, **kwargs):
"""Convert conversion results to a Dataset."""
measure = measure.lower()
y = self.get(measure)
v = self.get("v_{}".format(measure), error=False)
try:
n = self.get("n")
except:
n = None
return Dataset(y=y, v=v, n=n, **kwargs)
def test_2d_loop_warning(dataset_2d):
est = VarianceBasedLikelihoodEstimator()
y = np.random.normal(size=(10, 100))
v = np.random.randint(1, 50, size=(10, 100))
dataset = Dataset(y, v)
# Warning is raised when 2nd dim is > 10
with pytest.warns(UserWarning, match='Input contains'):
est.fit(dataset)
# But not when it's smaller
est.fit(dataset_2d)
def test_stouffers_perm_test_exact():
"""Test the exact permutation test on Stouffers Estimator."""
dataset = Dataset([1, 1, 2, 1.3], [1.5, 1, 2, 4])
results = StoufferCombinationTest().fit_dataset(dataset).summary()
pmr = results.permutation_test(2000)
assert pmr.n_perm == 16
assert pmr.exact
assert isinstance(pmr.results, CombinationTestResults)
assert pmr.perm_p["fe_p"].shape == (1, )
assert "tau2_p" not in pmr.perm_p
def test_exact_perm_test_1d_no_mods():
"""Test the exact permutation test on 1D data."""
dataset = Dataset([1, 1, 2, 1.3], [1.5, 1, 2, 4])
results = DerSimonianLaird().fit_dataset(dataset).summary()
pmr = results.permutation_test(867)
assert pmr.n_perm == 16
assert pmr.exact
assert isinstance(pmr.results, MetaRegressionResults)
assert pmr.perm_p["fe_p"].shape == (1, 1)
assert pmr.perm_p["tau2_p"].shape == (1, )
def test_stouffers_perm_test_approx():
"""Test the approximate permutation test on Stouffers Estimator."""
y = [2.8, -0.2, -1, 4.5, 1.9, 2.38, 0.6, 1.88, -0.4, 1.5, 3.163, 0.7]
dataset = Dataset(y)
results = StoufferCombinationTest().fit_dataset(dataset).summary()
pmr = results.permutation_test(2000)
assert not pmr.exact
assert pmr.n_perm == 2000
assert isinstance(pmr.results, CombinationTestResults)
assert pmr.perm_p["fe_p"].shape == (1, )
assert "tau2_p" not in pmr.perm_p
def test_dataset_init(variables):
dataset = Dataset(*variables, names=['bork'])
# Convenience accessors
assert np.array_equal(dataset.X, dataset.predictors)
assert np.array_equal(dataset.y, dataset.estimates)
assert np.array_equal(dataset.v, dataset.variances)
n = len(variables[0])
assert dataset.X.shape == (n, 2)
assert dataset.names == ['intercept', 'bork']
dataset = Dataset(*variables,
names=['bork'],
add_intercept=False,
extra_arg=200)
assert dataset.X.shape == (n, 1)
assert dataset.names == ['bork']
assert 'extra_arg' in dataset.kwargs
assert dataset.extra_arg == 200
def test_dataset_init_from_df(variables):
df = pd.DataFrame({
'y': [2, 4, 6],
'v_alt': [100, 100, 100],
'X1': [5, 2, 1],
'X7': [9, 8, 7]
})
dataset = Dataset(v='v_alt', X=['X1', 'X7'], data=df)
assert dataset.X.shape == (3, 3)
assert dataset.X_names == ['intercept', 'X1', 'X7']
assert np.array_equal(dataset.y, np.array([[2, 4, 6]]).T)
assert np.array_equal(dataset.v, np.array([[100, 100, 100]]).T)
def test_dataset_init_from_df(variables):
df = pd.DataFrame({
"y": [2, 4, 6],
"v_alt": [100, 100, 100],
"X1": [5, 2, 1],
"X7": [9, 8, 7]
})
dataset = Dataset(v="v_alt", X=["X1", "X7"], data=df)
assert dataset.X.shape == (3, 3)
assert dataset.X_names == ["intercept", "X1", "X7"]
assert np.array_equal(dataset.y, np.array([[2, 4, 6]]).T)
assert np.array_equal(dataset.v, np.array([[100, 100, 100]]).T)
def test_combination_test_from_dataset(Cls, data, mode, expected):
dset = Dataset(y=data)
est = Cls(mode).fit_dataset(dset)
results = est.summary()
z = ss.norm.isf(results.p)
assert np.allclose(z, expected, atol=1e-5)
def dataset(variables):
return Dataset(*variables, names=['my_covariate'])
def small_dataset_2d(variables):
"""Build a small Dataset with 2D data."""
y = np.array([[1.5, 1.9, 2.2], [4, 2, 1]]).T
v = np.array([[1, 0.8, 3], [1, 1.5, 1]]).T
return Dataset(y, v)
def dataset(variables):
"""Build a Dataset compiled from the variables fixture."""
return Dataset(*variables, X_names=["my_covariate"])
def dataset(variables):
return Dataset(*variables, X_names=["my_covariate"])
def small_dataset_2d(variables):
y = np.array([[1.5, 1.9, 2.2], [4, 2, 1]]).T
v = np.array([[1, 0.8, 3], [1, 1.5, 1]]).T
return Dataset(y, v)