def test_continuous_regression_with_overlap():
"""Test regression with overlap correction."""
signal = np.zeros(100000)
times = [1000, 2500, 3000, 5000, 5250, 7000, 7250, 8000]
events = np.zeros((len(times), 3), int)
events[:, 2] = 1
events[:, 0] = times
signal[events[:, 0]] = 1.
effect = hann(101)
signal = np.convolve(signal, effect)[:len(signal)]
raw = RawArray(signal[np.newaxis, :], mne.create_info(1, 100, 'eeg'))
assert_allclose(
effect,
linear_regression_raw(raw, events, {1: 1}, tmin=0)[1].data.flatten())
# test that sklearn solvers can be used
from sklearn.linear_model import ridge_regression
def solver(X, y):
return ridge_regression(X, y, alpha=0., solver="cholesky")
assert_allclose(
effect,
linear_regression_raw(raw, events, tmin=0,
solver=solver)['1'].data.flatten())
# test bad solvers
def solT(X, y):
return ridge_regression(X, y, alpha=0., solver="cholesky").T
pytest.raises(ValueError, linear_regression_raw, raw, events, solver=solT)
pytest.raises(ValueError, linear_regression_raw, raw, events, solver='err')
pytest.raises(TypeError, linear_regression_raw, raw, events, solver=0)
def test_continuous_regression_with_overlap():
"""Test regression with overlap correction."""
signal = np.zeros(100000)
times = [1000, 2500, 3000, 5000, 5250, 7000, 7250, 8000]
events = np.zeros((len(times), 3), int)
events[:, 2] = 1
events[:, 0] = times
signal[events[:, 0]] = 1.
effect = hann(101)
signal = np.convolve(signal, effect)[:len(signal)]
raw = RawArray(signal[np.newaxis, :], mne.create_info(1, 100, 'eeg'))
assert_allclose(effect, linear_regression_raw(
raw, events, {1: 1}, tmin=0)[1].data.flatten())
# test that sklearn solvers can be used
from sklearn.linear_model.ridge import ridge_regression
def solver(X, y):
return ridge_regression(X, y, alpha=0.)
assert_allclose(effect, linear_regression_raw(
raw, events, tmin=0, solver=solver)['1'].data.flatten())
# test bad solvers
def solT(X, y):
return ridge_regression(X, y, alpha=0.).T
assert_raises(ValueError, linear_regression_raw, raw, events, solver=solT)
assert_raises(ValueError, linear_regression_raw, raw, events, solver='err')
assert_raises(TypeError, linear_regression_raw, raw, events, solver=0)
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data."""
tmin, tmax = -.1, .5
raw = mne.io.read_raw_fif(raw_fname, preload=True)
raw.apply_proj()
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw, events, event_id, tmin, tmax,
baseline=None, reject=None)
revokeds = linear_regression_raw(raw, events, event_id,
tmin=tmin, tmax=tmax,
reject=None)
# Check that evokeds and revokeds are nearly equivalent
for cond in event_id.keys():
assert_allclose(revokeds[cond].data,
epochs[cond].average().data, rtol=1e-15)
# Test events that will lead to "duplicate" errors
old_latency = events[1, 0]
events[1, 0] = events[0, 0]
assert_raises(ValueError, linear_regression_raw,
raw, events, event_id, tmin, tmax)
events[1, 0] = old_latency
events[:, 0] = range(len(events))
assert_raises(ValueError, linear_regression_raw, raw,
events, event_id, tmin, tmax, decim=2)
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data"""
tmin, tmax = -.1, .5
raw = mne.io.Raw(raw_fname, preload=True)
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw,
events,
event_id,
tmin,
tmax,
baseline=None,
reject=None)
revokeds = linear_regression_raw(raw,
events,
event_id,
tmin=tmin,
tmax=tmax,
reject=None)
for cond in event_id.keys():
assert_allclose(revokeds[cond].data, epochs[cond].average().data)
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data"""
tmin, tmax = -.1, .5
raw = mne.io.Raw(raw_fname, preload=True)
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw, events, event_id, tmin, tmax,
baseline=None, reject=None)
revokeds = linear_regression_raw(raw, events, event_id,
tmin=tmin, tmax=tmax,
reject=None)
# Check that evokeds and revokeds are nearly equivalent
for cond in event_id.keys():
assert_allclose(revokeds[cond].data,
epochs[cond].average().data)
# Test events that will lead to "duplicate" errors
old_latency = events[1, 0]
events[1, 0] = events[0, 0]
assert_raises(ValueError, linear_regression_raw,
raw, events, event_id, tmin, tmax)
events[1, 0] = old_latency
events[:, 0] = range(len(events))
assert_raises(ValueError, linear_regression_raw, raw,
events, event_id, tmin, tmax, decim=2)
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data."""
tmin, tmax = -.1, .5
raw = mne.io.read_raw_fif(raw_fname, preload=True)
raw.apply_proj()
# a sampling of frequency where rounding and truncation yield
# different results checks conversion from samples to times is
# consistent across Epochs and linear_regression_raw
with raw.info._unlock():
raw.info['sfreq'] = 128
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw,
events,
event_id,
tmin,
tmax,
baseline=None,
reject=None)
revokeds = linear_regression_raw(raw,
events,
event_id,
tmin=tmin,
tmax=tmax,
reject=None)
# Check that evokeds and revokeds are nearly equivalent
for cond in event_id.keys():
assert_allclose(revokeds[cond].data,
epochs[cond].average().data,
rtol=1e-15)
# Test events that will lead to "duplicate" errors
old_latency = events[1, 0]
events[1, 0] = events[0, 0]
pytest.raises(ValueError, linear_regression_raw, raw, events, event_id,
tmin, tmax)
events[1, 0] = old_latency
events[:, 0] = range(len(events))
pytest.raises(ValueError,
linear_regression_raw,
raw,
events,
event_id,
tmin,
tmax,
decim=2)
def test_continuous_regression_with_overlap():
"""Test regression with overlap correction"""
signal = np.zeros(100000)
times = [1000, 2500, 3000, 5000, 5250, 7000, 7250, 8000]
events = np.zeros((len(times), 3), int)
events[:, 2] = 1
events[:, 0] = times
signal[events[:, 0]] = 1.
effect = hann(101)
signal = np.convolve(signal, effect)[:len(signal)]
raw = RawArray(signal[np.newaxis, :], mne.create_info(1, 100, 'eeg'))
assert_allclose(effect,
linear_regression_raw(raw, events, {1: 1}, tmin=0)[1]
.data.flatten())
def erp_regression(raw, events, event_id, tmin, tmax, reject, baseline):
"""This function processes the linear least-squares fitting of the
event-related potential of raw data
Estimation of ERP compared to actual ERP"""
from mne.stats.regression import linear_regression_raw
# standard epoching
epochs = mne.Epochs(raw,
events,
event_id,
tmin,
tmax,
reject=None,
baseline=None,
preload=True)
epochs.resample(150, npad='auto', n_jobs=1)
evokeds = linear_regression_raw(raw,
events=events,
event_id=event_id,
reject=None,
tmin=tmin,
tmax=tmax)
# linear_regression_raw returns a dict of evokeds
# select a condition similar to mne.Epochs objects
cond = raw_input(
"Please input the event ID you wish to plot the regression for: ")
if cond in event_id == True:
fig, (ax1, ax2, ax3) = fig.subplots(3, 1)
params = dict(spatial_colors=True,
show=False,
ylim=dict(eeg=(-10, 10)))
epochs[cond].average().plot(axes=ax1, **params)
evokeds[cond].plot(axes=ax2, **params)
contrast = mne.combine_evoked([evokeds[cond], -epochs[cond].average()],
weights='equal')
contrast.plot(axes=ax3, **params)
ax1.set_title("Traditional averaging")
ax2.set_title("rERF")
ax3.set_title("Difference")
plt.show()
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data"""
tmin, tmax = -.1, .5
raw = mne.io.Raw(raw_fname, preload=True)
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw, events, event_id, tmin, tmax,
baseline=None, reject=None)
revokeds = linear_regression_raw(raw, events, event_id,
tmin=tmin, tmax=tmax,
reject=None)
for cond in event_id.keys():
assert_allclose(revokeds[cond].data,
epochs[cond].average().data)
def test_continuous_regression_no_overlap():
"""Test regression without overlap correction, on real data."""
tmin, tmax = -.1, .5
raw = mne.io.read_raw_fif(raw_fname, preload=True)
raw.apply_proj()
# a sampling of frequency where rounding and truncation yield
# different results checks conversion from samples to times is
# consistent across Epochs and linear_regression_raw
raw.info['sfreq'] = 128
events = mne.read_events(event_fname)
event_id = dict(audio_l=1, audio_r=2)
raw = raw.pick_channels(raw.ch_names[:2])
epochs = mne.Epochs(raw, events, event_id, tmin, tmax,
baseline=None, reject=None)
revokeds = linear_regression_raw(raw, events, event_id,
tmin=tmin, tmax=tmax,
reject=None)
# Check that evokeds and revokeds are nearly equivalent
for cond in event_id.keys():
assert_allclose(revokeds[cond].data,
epochs[cond].average().data, rtol=1e-15)
# Test events that will lead to "duplicate" errors
old_latency = events[1, 0]
events[1, 0] = events[0, 0]
pytest.raises(ValueError, linear_regression_raw,
raw, events, event_id, tmin, tmax)
events[1, 0] = old_latency
events[:, 0] = range(len(events))
pytest.raises(ValueError, linear_regression_raw, raw,
events, event_id, tmin, tmax, decim=2)
tmax=tmax,
use_good_sensors=False)
# save_epochs_as_txt(epochs, os.path.join('pics', basename))
fig, axes = plt.subplots(1, 1)
raw.plot_sensors(show_names=True, axes=axes, show=False)
fig.savefig(os.path.join('pics', basename + '_sensor.png'), dpi=300)
fig, axes = plt.subplots(3, 1)
plot_tasks_in_timeline(raw, epochs, axes=axes)
fig.savefig(os.path.join('pics', basename + '_task.png'), dpi=600)
evoked = epochs.average()
evokeds = linear_regression_raw(raw,
events=epochs.events,
event_id=epochs.event_id,
reject=None,
tmin=tmin,
tmax=tmax)
for e in epochs.event_id.keys():
print(e)
fig, axes = plt.subplots(2, 1)
fig, title = plot_evoked(evokeds[e], axes=axes[0:2], title=basename)
fig.savefig(os.path.join('pics', title + e + '_ts_.png'), dpi=300)
fig, axes = plt.subplots(2, 1)
fig, title = plot_evoked(evoked, axes=axes[0:2], title=basename)
fig.savefig(os.path.join('pics', title + '_ts_.png'), dpi=300)
times = np.arange(0, 0.5, 0.05)
fig, axes = plt.subplots(6, times.shape[0])
idx = 0
for e in epochs.event_id.keys():
epochs = mne.Epochs(raw,
events,
event_id,
tmin,
tmax,
reject=None,
baseline=None,
preload=True,
verbose=False,
decim=4)
# rERF
evokeds = linear_regression_raw(raw,
events=events,
event_id=event_id,
reject=None,
tmin=tmin,
tmax=tmax,
decim=4)
# linear_regression_raw returns a dict of evokeds
# select conditions similarly to mne.Epochs objects
# plot both results
cond = "faces"
fig, (ax1, ax2) = plt.subplots(1, 2)
epochs[cond].average().plot(axes=ax1, show=False)
evokeds[cond].plot(axes=ax2, show=False)
ax1.set_title("Traditional averaging")
ax2.set_title("rERF")
plt.show()
raw = mne.io.Raw(raw_fname, preload=True) # Take first run
picks = mne.pick_types(raw.info, meg=True, exclude='bads')
raw.filter(1, 45, method='iir')
events = mne.find_events(raw, stim_channel='UPPT001')
event_id = dict(faces=1, scrambled=2)
tmin, tmax = -.1, .5
raw.pick_types(meg=True)
# regular epoching
epochs = mne.Epochs(raw, events, event_id, tmin, tmax, reject=None,
baseline=None, preload=True, verbose=False, decim=4)
# rERF
evokeds = linear_regression_raw(raw, events=events, event_id=event_id,
reject=None, tmin=tmin, tmax=tmax,
decim=4)
# linear_regression_raw returns a dict of evokeds
# select conditions similarly to mne.Epochs objects
# plot both results
cond = "faces"
fig, (ax1, ax2) = plt.subplots(1, 2)
epochs[cond].average().plot(axes=ax1, show=False)
evokeds[cond].plot(axes=ax2, show=False)
ax1.set_title("Traditional averaging")
ax2.set_title("rERF")
plt.show()
(RG_UB_events['A_index'] > 0).values
).astype(int)
RG_UB_events.loc[:, 'Intercept'] = 1
RG_UB_keys = [
s + '_' + 'IDX' + str(index_splits.index(idx)) for s in
['RG_UB_Auditory', 'RG_UB_Visual', 'RG_UB_Interaction']
]
RG_UB_events2 = RG_UB_events[[
"stim_onset", "button_press", "Intercept"
]].astype(int).values
RG_UB_covariates = RG_UB_events[RG_UB_keys]
UB_RG = linear_regression_raw(raw_tsss_mc,
events=RG_UB_events2,
event_id={},
covariates=RG_UB_covariates,
tmin=-0.15,
tmax=1.,
reject=None)
REG_list.extend([UB_RG[cond] for cond in RG_UB_keys])
RG_LB_events = pd.concat([LB_A, LB_V, LB_AVC])
RG_LB_events.loc[:, 'RG_LB_Auditory' + '_' + 'IDX' +
str(index_splits.index(idx))] = (
RG_LB_events['A_index'] >
0).astype(int).values
RG_LB_events.loc[:, 'RG_LB_Visual' + '_' + 'IDX' +
str(index_splits.index(idx))] = (
RG_LB_events['V_index'] >
0).astype(int).values
RG_LB_events.loc[:, 'RG_LB_Interaction' + '_' + 'IDX' +