def get_Tpermutation_distribution(dat_all, nperm, timeAll, labels):
time_start = timeAll[0][0]
time_end = timeAll[-1][1]
dat = dat_all[:, :,
int(time_start * 1000) + 100:int(time_end * 1000) + 100]
# permuta the data and calculate the largest sumed p values for each permutation across labels
highest_permutation_values = []
sign = np.ones((dat.shape[0], ))
sign[:np.int32(np.round(len(sign) / 2.))] = -1
for index in range(nperm): #to permute the conditions
print("permution: " + str(index))
sign = np.random.permutation(sign)
dat_perm = sign * np.transpose(dat, [1, 2, 0])
stats_perm, pval_perm = ttest_1samp(dat_perm, 0, axis=2)
stc_perm = SourceEstimate(
-np.log10(pval_perm),
vertices=[np.arange(10242), np.arange(10242)],
tmin=time_start,
tstep=1 / 1000.,
subject='fsaverage')
means_of_thresholded_timewinds_labels = []
for time1, time2 in timeAll:
stc_perm_timewind = stc_perm.copy().crop(time1, time2)
for label in labels:
label_fname = op.join(
'/autofs/cluster/kuperberg/nonconMM/MEG/MNE/250Labels',
label + '.label')
label_name = mne.read_label(label_fname,
subject='fsaverage',
color='r')
stc_label_perm = stc_perm_timewind.in_label(label_name)
stc_label_perm = threshold_by_pvalue(stc_label_perm, 2)
means_of_thresholded_timewinds_labels.append(
stc_label_perm.data.mean())
highest_permutation_values.append(
max(means_of_thresholded_timewinds_labels))
return highest_permutation_values
