def test_bootstrap_difference_and_fdr(self):
# Generate reference data
np.random.seed(31415)
x, var0 = self.generate_data()
a = cs.bootstrap_connectivity('PDC', x, VAR(2), 4, repeats=100)
# Similar to reference data ==> no significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC', x, VAR(2), 4, repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertFalse(np.any(p < 0.01))
self.assertFalse(np.any(cs.significance_fdr(p, 0.05)))
# Trials rearranged ==> no significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC',
x[:, :, ::-1],
VAR(2),
4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertFalse(np.any(p < 0.01))
self.assertFalse(np.any(cs.significance_fdr(p, 0.05)))
# Channels rearranged ==> highly significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC',
x[:, ::-1, 1],
VAR(2),
4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertTrue(np.all(p < 0.0001))
self.assertTrue(np.all(cs.significance_fdr(p, 0.01)))
# Time reversed ==> highly significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC',
x[::-1, :, 1],
VAR(2),
4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertTrue(np.all(p < 0.0001))
self.assertTrue(np.all(cs.significance_fdr(p, 0.01)))
def test_bootstrap(self):
np.random.seed(31415)
x, var0 = self.generate_data()
result = cs.bootstrap_connectivity('PDC', x, VAR(2), 4, repeats=100)
self.assertEqual(result.shape, (100, 2, 2, 4))
structure = np.mean(np.mean(result, axis=3), axis=0)
# make sure result has roughly the correct structure
self.assertTrue(np.all(np.abs(structure - [[1, 0], [0.5, 1]]) < 0.25))
def test_bootstrap(self):
np.random.seed(31415)
x, var0 = self.generate_data()
result = cs.bootstrap_connectivity('PDC', x, VAR(2), nfft=4,
repeats=100)
self.assertEqual(result.shape, (100, 2, 2, 4))
structure = np.mean(np.mean(result, axis=3), axis=0)
# make sure result has roughly the correct structure
self.assertTrue(np.all(np.abs(structure - [[1, 0], [0.5, 1]]) < 0.25))
def test_bootstrap_difference_and_fdr(self):
# Generate reference data
np.random.seed(31415)
x, var0 = self.generate_data()
a = cs.bootstrap_connectivity('PDC', x, VAR(2), nfft=4, repeats=100)
# Similar to reference data ==> no significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC', x, VAR(2), nfft=4, repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertFalse(np.any(p < 0.01)) # TODO: np.all?
self.assertFalse(np.any(cs.significance_fdr(p, 0.05))) # TODO: np.all?
# Trials rearranged ==> no significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC', x[::-1, :, :], VAR(2), nfft=4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertFalse(np.any(p < 0.01))
self.assertFalse(np.any(cs.significance_fdr(p, 0.05)))
# Channels rearranged ==> highly significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC', x[1, ::-1, :], VAR(2), nfft=4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertTrue(np.all(p < 0.0001))
self.assertTrue(np.all(cs.significance_fdr(p, 0.01)))
# Time reversed ==> highly significant differences expected
np.random.seed(12345)
x, var0 = self.generate_data()
b = cs.bootstrap_connectivity('PDC', x[1, :, ::-1], VAR(2), nfft=4,
repeats=100)
p = cs.test_bootstrap_difference(a, b)
self.assertTrue(np.all(p < 0.0001))
self.assertTrue(np.all(cs.significance_fdr(p, 0.01)))
