def test_f_twoway_rm():
""" Test 2-way anova """
rng = np.random.RandomState(42)
iter_params = product([4, 10], [2, 15], [4, 6, 8], ['A', 'B', 'A:B'],
[False, True])
_effects = {4: [2, 2], 6: [2, 3], 8: [2, 4]}
for params in iter_params:
n_subj, n_obs, n_levels, effects, correction = params
data = rng.random_sample([n_subj, n_levels, n_obs])
fvals, pvals = f_mway_rm(data,
_effects[n_levels],
effects,
correction=correction)
assert_true((fvals >= 0).all())
if pvals.any():
assert_true(((0 <= pvals) & (1 >= pvals)).all())
n_effects = len(_map_effects(n_subj, effects)[0])
assert_true(fvals.size == n_obs * n_effects)
if n_effects == 1: # test for principle of least surprise ...
assert_true(fvals.ndim == 1)
fvals_ = f_threshold_mway_rm(n_subj, _effects[n_levels], effects)
assert_true((fvals_ >= 0).all())
assert_true(fvals_.size == n_effects)
data = rng.random_sample([n_subj, n_levels, 1])
assert_raises(ValueError,
f_mway_rm,
data,
_effects[n_levels],
effects='C',
correction=correction)
data = rng.random_sample([n_subj, n_levels, n_obs, 3])
# check for dimension handling
f_mway_rm(data, _effects[n_levels], effects, correction=correction)
# now check against external software results
test_data = generate_data(n_subjects=20, n_conditions=6)
fvals, pvals = f_mway_rm(test_data, [2, 3])
assert_array_almost_equal(fvals, test_external['spss_fvals'], 3)
assert_array_almost_equal(pvals, test_external['spss_pvals_uncorrected'],
3)
assert_array_almost_equal(fvals, test_external['r_fvals'], 4)
assert_array_almost_equal(pvals, test_external['r_pvals_uncorrected'], 3)
_, pvals = f_mway_rm(test_data, [2, 3], correction=True)
assert_array_almost_equal(pvals, test_external['spss_pvals_corrected'], 3)
test_data = generate_data(n_subjects=20, n_conditions=8)
fvals, _ = f_mway_rm(test_data, [2, 2, 2])
assert_array_almost_equal(fvals, test_external['r_fvals_3way'], 5)
fvals, _ = f_mway_rm(test_data, [8], 'A')
assert_array_almost_equal(fvals, test_external['r_fvals_1way'], 5)
def test_f_twoway_rm():
"""Test 2-way anova."""
rng = np.random.RandomState(42)
iter_params = product([4, 10], [2, 15], [4, 6, 8],
['A', 'B', 'A:B'],
[False, True])
_effects = {
4: [2, 2],
6: [2, 3],
8: [2, 4]
}
for params in iter_params:
n_subj, n_obs, n_levels, effects, correction = params
data = rng.random_sample([n_subj, n_levels, n_obs])
fvals, pvals = f_mway_rm(data, _effects[n_levels], effects,
correction=correction)
assert (fvals >= 0).all()
if pvals.any():
assert ((0 <= pvals) & (1 >= pvals)).all()
n_effects = len(_map_effects(n_subj, effects)[0])
assert fvals.size == n_obs * n_effects
if n_effects == 1: # test for principle of least surprise ...
assert fvals.ndim == 1
fvals_ = f_threshold_mway_rm(n_subj, _effects[n_levels], effects)
assert (fvals_ >= 0).all()
assert fvals_.size == n_effects
data = rng.random_sample([n_subj, n_levels, 1])
pytest.raises(ValueError, f_mway_rm, data, _effects[n_levels],
effects='C', correction=correction)
data = rng.random_sample([n_subj, n_levels, n_obs, 3])
# check for dimension handling
f_mway_rm(data, _effects[n_levels], effects, correction=correction)
# now check against external software results
test_data = generate_data(n_subjects=20, n_conditions=6)
fvals, pvals = f_mway_rm(test_data, [2, 3])
assert_array_almost_equal(fvals, test_external['spss_fvals'], 3)
assert_array_almost_equal(pvals, test_external['spss_pvals_uncorrected'],
3)
assert_array_almost_equal(fvals, test_external['r_fvals'], 4)
assert_array_almost_equal(pvals, test_external['r_pvals_uncorrected'], 3)
_, pvals = f_mway_rm(test_data, [2, 3], correction=True)
assert_array_almost_equal(pvals, test_external['spss_pvals_corrected'], 3)
test_data = generate_data(n_subjects=20, n_conditions=8)
fvals, _ = f_mway_rm(test_data, [2, 2, 2])
assert_array_almost_equal(fvals, test_external['r_fvals_3way'], 5)
fvals, _ = f_mway_rm(test_data, [8], 'A')
assert_array_almost_equal(fvals, test_external['r_fvals_1way'], 5)