def test_signature_significance_fdr_bh_corr(self):
"""
A typical corr='fdr_bh' scenario, that requires statsmodels.
"""
sig_spectrum = spade.test_signature_significance(
pv_spec=((2, 3, 0.2), (2, 4, 0.05)),
concepts=([[(2, 3), (1, 2, 3)],
[(2, 4), (1, 2, 3, 4)]]),
alpha=0.15, winlen=1, corr='fdr_bh')
self.assertEqual(sig_spectrum, [(2., 3., False), (2., 4., True)])
config = yaml.load(stream)
psr_param = config['psr_param']
correction = config['correction']
path = '../results/results_data{0}/'.format(data_idx)
res_ind, params = np.load(path + 'result_ind.npy', encoding='latin1')
pval_spectrum = res_ind[0]['pvalue_spectrum']
spectrum = params['spectrum']
if len(pval_spectrum) == 0:
ns_sgnt = []
else:
# Computing non-significant entries of the spectrum applying
# the statistical correction
ns_sgnt = spade.test_signature_significance(pval_spectrum,
alpha,
corr=correction,
report='non_significant',
spectrum=spectrum)
filtered_results = {}
for xi in range(2, 11):
for occ in range(3, 11):
try:
results, params = np.load(
path + 'result_xi{xi}_occ{occ}.npy'.format(xi=xi, occ=occ),
encoding='latin1')
except:
print('not all res available')
continue
filtered_results['xi{xi}_occ{occ}'.format(xi=xi, occ=occ)] = {}
# SPADE parameters