def test_plot_windows(self):
"""Test function plot_windows."""
n_pts = 1000
times = np.linspace(-1, 1, n_pts, endpoint=True)
kw = dict(verbose=False)
ts = define_windows(times, slwin_len=.1, **kw)[0]
plot_windows(times, ts)
def test_define_windows(self):
"""Test function define_windows."""
n_pts = 1000
times = np.linspace(-1, 1, n_pts, endpoint=True)
kw = dict(verbose=False)
# ---------------------------------------------------------------------
# full window
# ---------------------------------------------------------------------
ts = define_windows(times, **kw)[0]
np.testing.assert_array_equal(ts, np.array([[0, n_pts - 1]]))
# ---------------------------------------------------------------------
# custom windows
# ---------------------------------------------------------------------
# single window
win = [0, .5]
ts = define_windows(times, windows=win, **kw)[0]
np.testing.assert_almost_equal(times[ts.ravel()], win, decimal=2)
# multiple
win = [[-.5, -.1], [-.1, 0.], [0, .5]]
ts = define_windows(times, windows=win, **kw)[0]
np.testing.assert_almost_equal(times[ts], np.array(win), decimal=2)
# ---------------------------------------------------------------------
# sliding windows
# ---------------------------------------------------------------------
# length only
ts = define_windows(times, slwin_len=.1, **kw)[0]
tts = times[ts]
ttsd = np.diff(tts, axis=1)
np.testing.assert_almost_equal(ttsd, np.full_like(ttsd, .1), decimal=2)
# with starting point
ts = define_windows(times, slwin_len=.1, slwin_start=.5, **kw)[0]
np.testing.assert_almost_equal(times[ts][0, 0], .5, decimal=2)
# with stoping point
ts = define_windows(times, slwin_len=.1, slwin_stop=.9, **kw)[0]
assert times[ts][-1, -1] <= .9
# with step between temporal windows
ts = define_windows(times, slwin_len=.1, slwin_step=.2, **kw)[0]
tts = times[ts]
ttsd = np.diff(tts, axis=0)
np.testing.assert_almost_equal(ttsd, np.full_like(ttsd, .2), decimal=2)
times = np.linspace(-1, 1.5, n_pts, endpoint=True) x = np.sin(2 * np.pi * (1. / period) * times) ############################################################################### # Manually define windows # ----------------------- # # The first feature we illustrate here, is to define windows manually which # means, based on the time vector, where each window should start and finish. plt.figure(figsize=(10, 8)) # by default, if you don't provide any input, the full time window is going to # be considered win = define_windows(times)[0] plt.subplot(311) plot_windows(times, win, x, title='No input = full time window') # however, you can also specify where each window start and finish. Here, we # define a simple unique window between [-.5, .5] win = define_windows(times, windows=[-.5, .5])[0] plt.subplot(312) plot_windows(times, win, x, title='Unique window [-.5, .5]') # but you can also specify multiple windows win = define_windows(times, windows=[[-.75, -.25], [.25, .75]])[0] plt.subplot(313) plot_windows(times, win, x, title='Multiple windows manually defined') plt.tight_layout() plt.show()
###############################################################################
# Stimulus-specificity of the undirected connectivity
# ---------------------------------------------------
#
# From the figure above, nodes X and Y present an activity that is modulated
# according to the stimulus, but not Z. The next question we can ask is
# whether the connectivity strength is also modulated by the stimulus. To this
# end, we are going to compute the undirected Dynamic Functional Connectivity
# (DFC) which is simply defined as the information shared between two nodes
# inside sliding windows. Hence, here, the DFC is computed for each trial
# inside consecutive windows
# define the sliding windows
slwin_len = .3 # 100ms window length
slwin_step = .02 # 80ms between consecutive windows
win_sample = define_windows(times, slwin_len=slwin_len,
slwin_step=slwin_step)[0]
# compute the DFC for each subject
dfc = []
for n_s in range(n_subjects):
_dfc = conn_dfc(x[n_s].data, win_sample, times=times, roi=roi,
verbose=False)
# reset trials dimension
_dfc['trials'] = x[n_s]['trials'].data
dfc += [_dfc]
###############################################################################
# now we can plot the dfc by concatenating all of the subjects
dfc_suj = xr.concat(dfc, 'trials').groupby('trials').mean('trials')
dfc_suj.plot.line(x='times', col='roi', hue='trials')