def create_dm(self, drop=None, convolve=True):
""" Create a unit (boxcar-only) DM with one columns for each
condition in self.trials.
If <convolve> the dm is convolved with the HRF (self.hrf). """
cond_levels = sorted(list(set(self.trials)))
## Find and sort conditions in trials
# Some useful counts...
num_conds = len(cond_levels)
num_tr = np.sum(self.durations)
# Map each condition in trials to a
# 2d binary 2d array. Each row is a trial
# and each column is a condition.
dm_unit = np.zeros((num_tr, num_conds))
for col, cond in enumerate(cond_levels):
# Create boolean array use it to
# populate the dm with ones...
# which must be in tr time.
mask_in_tr = dtime(self.trials == cond, self.durations, drop,
False)
dm_unit[mask_in_tr, col] = 1
self.dm = dm_unit
if convolve:
self.dm = self._convolve_hrf(self.dm)
def create_dm(self, drop=None, convolve=True):
""" Create a unit (boxcar-only) DM with one columns for each
condition in self.trials.
If <convolve> the dm is convolved with the HRF (self.hrf). """
cond_levels = sorted(list(set(self.trials)))
## Find and sort conditions in trials
# Some useful counts...
num_conds = len(cond_levels)
num_tr = np.sum(self.durations)
# Map each condition in trials to a
# 2d binary 2d array. Each row is a trial
# and each column is a condition.
dm_unit = np.zeros((num_tr, num_conds))
for col, cond in enumerate(cond_levels):
# Create boolean array use it to
# populate the dm with ones...
# which must be in tr time.
mask_in_tr = dtime(
self.trials == cond, self.durations, drop, False)
dm_unit[mask_in_tr,col] = 1
self.dm = dm_unit
if convolve:
self.dm = self._convolve_hrf(self.dm)
def create_dm_param(self,
names,
drop=None,
box=True,
orth=False,
convolve=True):
""" Create a parametric design matrix based on <names> in self.data.
If <box> a univariate dm is created that fills the leftmost
side of the dm.
If <orth> each regressor is orthgonalized with respect to its
left-hand neighbor (excluding the baseline).
If <convolve> the dm is convolved with the HRF (self.hrf). """
cond_levels = sorted(list(set(self.trials)))
## Find and sort conditions in trials
# Some useful counts...
num_names = len(names)
num_tr = np.sum(self.durations)
dm_param = None
## Will eventually hold the
## parametric DM.
for cond in cond_levels:
if cond == 0:
continue
## We add the baseline
## in at the end
# Create a temp dm to hold this
# condition"s data
dm_temp = np.zeros((num_tr, num_names))
mask_in_tr = dtime(self.trials == cond, self.durations, drop,
False)
# Get the named data, convert to tr time
# then add to the temp dm using the mask
for col, name in enumerate(names):
data_in_tr = dtime(self.data[name], self.durations, drop, 0)
dm_temp[mask_in_tr, col] = data_in_tr[mask_in_tr]
# Store the temporary DM in
# the final DM.
if dm_param == None:
dm_param = dm_temp ## reinit
else:
dm_param = np.hstack((dm_param, dm_temp)) ## adding
# Create the unit DM too, then combine them.
# defining self.dm in the process
self.create_dm(convolve=False)
dm_unit = self.dm.copy()
self.dm = None
## Copy and reset
if box:
self.dm = np.hstack((dm_unit, dm_param))
else:
baseline = dm_unit[:, 0]
baseline = baseline.reshape(baseline.shape[0], 1)
self.dm = np.hstack((baseline, dm_param))
## If not including the boxcar,
## we still need the baseline model.
# Orthgonalize the regessors?
if orth:
self._orth_dm()
# Convolve with self.hrf?
if convolve:
self.dm = self._convolve_hrf(self.dm)
def create_dm_param(self, names, drop=None, box=True, orth=False, convolve=True):
""" Create a parametric design matrix based on <names> in self.data.
If <box> a univariate dm is created that fills the leftmost
side of the dm.
If <orth> each regressor is orthgonalized with respect to its
left-hand neighbor (excluding the baseline).
If <convolve> the dm is convolved with the HRF (self.hrf). """
cond_levels = sorted(list(set(self.trials)))
## Find and sort conditions in trials
# Some useful counts...
num_names = len(names)
num_tr = np.sum(self.durations)
dm_param = None
## Will eventually hold the
## parametric DM.
for cond in cond_levels:
if cond == 0:
continue
## We add the baseline
## in at the end
# Create a temp dm to hold this
# condition"s data
dm_temp = np.zeros((num_tr, num_names))
mask_in_tr = dtime(
self.trials == cond, self.durations, drop, False)
# Get the named data, convert to tr time
# then add to the temp dm using the mask
for col, name in enumerate(names):
data_in_tr = dtime(
self.data[name], self.durations, drop, 0)
dm_temp[mask_in_tr,col] = data_in_tr[mask_in_tr]
# Store the temporary DM in
# the final DM.
if dm_param == None:
dm_param = dm_temp ## reinit
else:
dm_param = np.hstack((dm_param, dm_temp)) ## adding
# Create the unit DM too, then combine them.
# defining self.dm in the process
self.create_dm(convolve=False)
dm_unit = self.dm.copy()
self.dm = None
## Copy and reset
if box:
self.dm = np.hstack((dm_unit, dm_param))
else:
baseline = dm_unit[:,0]
baseline = baseline.reshape(baseline.shape[0], 1)
self.dm = np.hstack((baseline, dm_param))
## If not including the boxcar,
## we still need the baseline model.
# Orthgonalize the regessors?
if orth:
self._orth_dm()
# Convolve with self.hrf?
if convolve:
self.dm = self._convolve_hrf(self.dm)
