stat = 'beta' #choose whether you want to use the single subject tstats or betas for the analysis
if stat == 'beta':
dat2use = deepcopy(data)
elif stat == 'tstat':
dat2use = deepcopy(data_t)
#for channel in ['FCz', 'Cz', 'CPz', 'Pz', 'POz', 'Oz']:
for channel in ['FCz', 'Cz', 'Pz']:
tmin = 0
tmax = 1
#incorrect vs definitely correct
t_ivsd, clu_ivsd, clupv_ivsd, _ = runclustertest_epochs(data = dat2use,
contrast_name = 'incorrvsdef',
channels = [channel],
tmin = tmin,
tmax = tmax,
gauss_smoothing = None,
out_type = 'indices', n_permutations = 10000
)
clutimes = deepcopy(dat2use['incorrect'][0]).crop(tmin = tmin, tmax = tmax).times
masks_ivsd = np.asarray(clu_ivsd)[clupv_ivsd <= 0.05]
#incorrect vs just correct
t_ivsj, clu_ivsj, clupv_ivsj, _ = runclustertest_epochs(data = dat2use,
contrast_name = 'incorrvsjust',
channels = [channel],
tmin = tmin,
tmax = tmax,
gauss_smoothing = None,
out_type = 'indices', n_permutations = 10000
legend = 'upper right', picks = channel,
ci = .68, show_sensors = False, title = 'electrode ' + channel,
truncate_xaxis=False
)
gave = mne.grand_average(data['incorrvscorr'])
deepcopy(gave).plot_joint(times = np.arange(0.1,0.7,.1),
title = 'incorr vs corr')
#%%
tmin, tmax = 0, 1
for channel in ['FCz', 'Cz']:
t_ern, clu_ern, clupv_ern, h0_ern = runclustertest_epochs(data = data,
contrast_name = 'incorrvscorr',
channels = [channel],
tmin = tmin,
tmax = tmax,
gauss_smoothing = None,
out_type = 'indices', n_permutations = 'Default'
)
clutimes = deepcopy(data['grandmean'][0]).crop(tmin = tmin, tmax = tmax).times
masks_ern = np.asarray(clu_ern)[clupv_ern < 0.05]
fig = plt.figure()
ax = plt.axes()
mne.viz.plot_compare_evokeds(
evokeds = dict(
corr = data['correct'],
incorr = data['incorrect'],
diff = data['incorrvscorr']
),
title='incorrect vs just correct',
times=np.arange(0.1, 0.6, 0.05))
gave_justvsdef.plot_joint(picks='eeg',
topomap_args=dict(outlines='head', contours=0),
title='just correct vs definitely correct',
times=np.arange(0.1, 0.6, 0.05))
#%%
#nonpara cluster t test to see where diff is significant
tmin, tmax = 0, 1
channel = 'FCz'
t_ivsd, clu_ivsd, clupv_ivsd, H0_ivsd = runclustertest_epochs(
data=data,
channels=[channel],
contrast_name='incorrvsdef',
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
n_permutations='Default')
masks_ivsd = np.asarray(clu_ivsd)[clupv_ivsd <= 0.05]
t_ivsj, clu_ivsj, clupv_ivsj, H0_ivsj = runclustertest_epochs(
data=data,
channels=[channel],
contrast_name='incorrvsjust',
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
n_permutations='Default')
ci=.68,
show_sensors=False,
title='electrode ' + channel,
truncate_xaxis=False)
gave = mne.grand_average(data['errorincorrvscorr'])
deepcopy(gave).plot_joint(times=np.arange(0.1, 0.7, .1),
title='error in incorrect vs correct trials')
tmin, tmax = 0, 1
for channel in ['FCz', 'Cz']:
t_ern, clu_ern, clupv_ern, h0_ern = runclustertest_epochs(
data=data,
contrast_name='errorincorrect',
channels=[channel],
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
n_permutations='Default')
clutimes = deepcopy(data['grandmean'][0]).crop(tmin=tmin, tmax=tmax).times
masks_ern = np.asarray(clu_ern)[clupv_ern < 0.05]
fig = plt.figure()
ax = plt.axes()
mne.viz.plot_compare_evokeds(
evokeds=dict(
corr=data['errorcorrect'],
incorr=data['errorincorrect'],
#diff = data['errorincorrvscorr'],
all=data['error']),
for channel in ['FCz', 'Cz']:
dictobj[contrast][channel] = []
clutimes = deepcopy(data['correct'][0]).crop(tmin=tmin, tmax=tmax).times
for contrast in [
'confdiff_correct', 'confdiff_incorrect', 'confdiff_incorrvscorr',
'confdiff'
]:
# for contrast in ['confdiff']:
for channel in ['FCz', 'Cz']:
t_cope[contrast][channel], clu_cope[contrast][channel], clupv_cope[
contrast][channel], h0_cope[contrast][
channel] = runclustertest_epochs(data=data,
contrast_name=contrast,
channels=[channel],
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
n_permutations=5000)
masks_cope[contrast][channel] = np.asarray(
clu_cope[contrast][channel])[clupv_cope[contrast][channel] < 0.05]
for channel in ['FCz']:
fig = plt.figure()
ax = fig.subplots(1)
mne.viz.plot_compare_evokeds(
evokeds=dict(correct=data['confdiff_correct'],
incorrect=data['confdiff_incorrect'],
difference=data['confdiff_incorrvscorr'],
all_trials=data['confdiff']),
colors=dict(correct='#4daf4a',
#for channel in ['FCz', 'Cz', 'CPz', 'Pz', 'POz', 'Oz']:
#for channel in ['FCz', 'Cz', 'CPz', 'Pz', 'POz', 'Oz']:
for channel in ['FCz', 'Cz', 'Pz']:
tmin = 0
tmax = 1
clutimes = deepcopy(dat2use['incorrect'][0]).crop(tmin=tmin,
tmax=tmax).times
#three contrasts of interest here:
#incorrect vs definitely correct
t_ivsd, clu_ivsd, clupv_ivsd, _ = runclustertest_epochs(
data=dat2use,
contrast_name='errorincorrvsdef',
channels=[channel],
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
n_permutations=10000)
masks_ivsd = np.asarray(clu_ivsd)[clupv_ivsd <= 0.05]
#incorrect vs just correct
t_ivsj, clu_ivsj, clupv_ivsj, _ = runclustertest_epochs(
data=dat2use,
contrast_name='errorincorrvsjust',
channels=[channel],
tmin=tmin,
tmax=tmax,
gauss_smoothing=None,
out_type='indices',
ci = .68, show_sensors = False, title = 'electrode ' + channel,
truncate_xaxis=False
)
gave = mne.grand_average(data['incorrvscorr'])
deepcopy(gave).plot_joint(times = np.arange(0.1,0.7,.1),
title = 'incorr vs corr - RM reference',
picks='eeg')
tmin, tmax = 0, 1
for channel in ['FCZ', 'CZ', 'CPZ', 'PZ']:
t_ern, clu_ern, clupv_ern, h0_ern = runclustertest_epochs(data = data,
contrast_name = 'incorrvscorr',
channels = [channel],
tmin = tmin,
tmax = tmax,
gauss_smoothing = None,
out_type = 'indices', n_permutations = 'Default'
)
clutimes = deepcopy(data['grandmean'][0]).crop(tmin = tmin, tmax = tmax).times
masks_ern = np.asarray(clu_ern)[clupv_ern < 0.05]
fig = plt.figure()
ax = plt.axes()
mne.viz.plot_compare_evokeds(
evokeds = dict(
corr = data['correct'],
incorr = data['incorrect'],
diff = data['incorrvscorr']
),
show_sensors = False, title = 'electrode: '+channel, truncate_xaxis = False)
# this is reassuring there arent differences in ERP shape based on cueing at either the frontal electrodes of interest
# or at posterior visual (central because central stimulus) electrodes
tmin, tmax = 0,1 #set time window for the clustering
#gonna store things in a dictionary so we dont need to constantly rerun
t_ern, clu_ern, clupv_ern, h0_ern = dict(), dict(), dict(), dict()
masks_ern = dict()
clutimes = deepcopy(data['grandmean'][0].crop(tmin = tmin, tmax = tmax).times)
for channel in ['FCz', 'Cz','Pz', 'POz', 'Oz']:
t_ern[channel], clu_ern[channel], clupv_ern[channel], h0_ern[channel] = runclustertest_epochs(data = data,
contrast_name = 'cuedvsneutral',
channels = [channel],
tmin = tmin, tmax = tmax,
gauss_smoothing = None,
out_type = 'indices', n_permutations = 5000)
masks_ern[channel] = np.asarray(clu_ern[channel])[clupv_ern[channel] < 0.05]
#no clusters survive permutation testing at any electrode, for the difference between cued and neutral
#given this fact, having these regressors (neutral and cued categorical regressors) in the model
# only increases model complexity without better explaining the data in a meaningful way, so we can remove them
#%%
#first pass look at scalp topographies (joint plot) for neutral/cued trials and their difference
mne.grand_average(deepcopy(data['pside'])).plot_joint(times = np.arange(0.05, 0.7, 0.1), title = 'probed side (lvsr)')
#plot some posterior channels also incase there are posterior differences based on cueing
for channel in ['PO3', 'PO7', 'O1']:
mne.viz.plot_compare_evokeds(
