def xcorr_eta(self):
"""Compute the normalized cross-correlation estimate of the HRFs for
different kinds of events
Returns
-------
A time-series object, shape[:-2] are dimensions corresponding to the to
shape[:-2] of the EventRelatedAnalyzer data, shape[-2] corresponds to
the different kinds of events used (ordered according to the sorted
order of the unique components in the events time-series). shape[-1]
corresponds to time, and has length = len_et (xcorr looks both back
and forward for half of this length)
"""
#Make a list to put the outputs in:
h = [0] * self._len_h
for i in range(self._len_h):
data = self.data[i]
u = np.unique(self.events[i])
event_types = u[np.unique(self.events[i]) != 0]
h[i] = np.empty((event_types.shape[0],
self.len_et // 2),
dtype=complex)
for e_idx in range(event_types.shape[0]):
this_e = (self.events[i] == event_types[e_idx]) * 1.0
if self._zscore:
this_h = tsa.freq_domain_xcorr_zscored(
data,
this_e,
-self.offset + 1,
self.len_et - self.offset - 2)
else:
this_h = tsa.freq_domain_xcorr(
data,
this_e,
-self.offset + 1,
self.len_et - self.offset - 2)
h[i][e_idx] = this_h
h = np.array(h).squeeze()
## t0 for the object returned here needs to be the central time, not
## the first time point, because the functions 'look' back and forth
## for len_et bins
return ts.TimeSeries(data=h,
sampling_rate=self.sampling_rate,
t0=-1 * self.len_et * self.sampling_interval,
time_unit=self.time_unit)
def xcorr_eta(self):
"""Compute the normalized cross-correlation estimate of the HRFs for
different kinds of events
Returns
-------
A time-series object, shape[:-2] are dimensions corresponding to the to
shape[:-2] of the EventRelatedAnalyzer data, shape[-2] corresponds to
the different kinds of events used (ordered according to the sorted
order of the unique components in the events time-series). shape[-1]
corresponds to time, and has length = len_et (xcorr looks both back
and forward for half of this length)
"""
#Make a list to put the outputs in:
h = [0] * self._len_h
for i in range(self._len_h):
data = self.data[i]
u = np.unique(self.events[i])
event_types = u[np.unique(self.events[i]) != 0]
h[i] = np.empty((event_types.shape[0],
self.len_et // 2),
dtype=complex)
for e_idx in range(event_types.shape[0]):
this_e = (self.events[i] == event_types[e_idx]) * 1.0
if self._zscore:
this_h = tsa.freq_domain_xcorr_zscored(
data,
this_e,
-self.offset + 1,
self.len_et - self.offset - 2)
else:
this_h = tsa.freq_domain_xcorr(
data,
this_e,
-self.offset + 1,
self.len_et - self.offset - 2)
h[i][e_idx] = this_h
h = np.array(h).squeeze()
## t0 for the object returned here needs to be the central time, not
## the first time point, because the functions 'look' back and forth
## for len_et bins
return ts.TimeSeries(data=h,
sampling_rate=self.sampling_rate,
t0=-1 * self.len_et * self.sampling_interval,
time_unit=self.time_unit)
def test_xcorr_zscored():
"""
Test this function, which is not otherwise tested in the testing of the
EventRelatedAnalyzer
"""
cycles = 10
l = 1024
unit = 2 * np.pi / l
t = np.arange(0, 2 * np.pi + unit, unit)
signal = np.sin(cycles * t)
events = np.zeros(t.shape)
#Zero crossings:
idx = np.where(np.abs(signal) < 0.03)[0]
#An event occurs at the beginning of every cycle:
events[idx[:-2:2]] = 1
a = tsa.freq_domain_xcorr_zscored(signal, events, 1000, 1000)
npt.assert_almost_equal(np.mean(a), 0, 1)
npt.assert_almost_equal(np.std(a), 1, 1)
