def test_f_score_with_covars_and_normalized_design_withcovar(random_state=0):
"""
This test has a statsmodels dependance. There seems to be no simple,
alternative way to perform a F-test on a linear model including
covariates.
"""
try:
from statsmodels.regression.linear_model import OLS
except:
warnings.warn("Statsmodels is required to run this test")
raise nose.SkipTest
rng = check_random_state(random_state)
### Normalized data
n_samples = 50
# generate data
var1 = np.ones((n_samples, 1)) / np.sqrt(n_samples) # normalized
var2 = rng.randn(n_samples, 1)
var2 = var2 / np.sqrt(np.sum(var2**2, 0)) # normalize
covars = np.eye(n_samples, 3) # covars is orthogonal
covars[3] = -1 # covars is orthogonal to var1
covars = orthonormalize_matrix(covars)
# own f_score
f_val_own = _f_score_with_covars_and_normalized_design(var1, var2,
covars)[0]
# statsmodels f_score
test_matrix = np.array([[1., 0., 0., 0.]])
statsmodels_ols = OLS(var2, np.hstack((var1, covars))).fit()
f_val_statsmodels = statsmodels_ols.f_test(test_matrix).fvalue[0]
assert_array_almost_equal(f_val_own, f_val_statsmodels)
def test_f_score_with_covars_and_normalized_design_withcovar(random_state=0):
"""
This test has a statsmodels dependance. There seems to be no simple,
alternative way to perform a F-test on a linear model including
covariates.
"""
try:
from statsmodels.regression.linear_model import OLS
except:
warnings.warn("Statsmodels is required to run this test")
raise nose.SkipTest
rng = check_random_state(random_state)
### Normalized data
n_samples = 50
# generate data
var1 = np.ones((n_samples, 1)) / np.sqrt(n_samples) # normalized
var2 = rng.randn(n_samples, 1)
var2 = var2 / np.sqrt(np.sum(var2 ** 2, 0)) # normalize
covars = np.eye(n_samples, 3) # covars is orthogonal
covars[3] = -1 # covars is orthogonal to var1
covars = orthonormalize_matrix(covars)
# own f_score
f_val_own = _f_score_with_covars_and_normalized_design(var1, var2, covars)[0]
# statsmodels f_score
test_matrix = np.array([[1.0, 0.0, 0.0, 0.0]])
statsmodels_ols = OLS(var2, np.hstack((var1, covars))).fit()
f_val_statsmodels = statsmodels_ols.f_test(test_matrix).fvalue[0]
assert_array_almost_equal(f_val_own, f_val_statsmodels)
def test_t_score_with_covars_and_normalized_design_withcovar(random_state=0):
"""
"""
rng = check_random_state(random_state)
### Normalized data
n_samples = 50
# generate data
var1 = np.ones((n_samples, 1)) / np.sqrt(n_samples) # normalized
var2 = rng.randn(n_samples, 1)
var2 = var2 / np.sqrt(np.sum(var2 ** 2, 0)) # normalize
covars = np.eye(n_samples, 3) # covars is orthogonal
covars[3] = -1 # covars is orthogonal to var1
covars = orthonormalize_matrix(covars)
# nilearn t-score
own_score = _t_score_with_covars_and_normalized_design(var1, var2, covars)
# compute t-scores with linalg or statmodels
ref_score = get_tvalue_with_alternative_library(var1, var2, covars)
assert_array_almost_equal(own_score, ref_score)
