def calc_max_repeats(self):
'''
Returns maximum number of repeats of any one task in the dataset.
Determines the shape of the sorted data arrays.
Returns
-------
max_reps : int
maximum number of repeats of any one direction in center out
'''
n_trials = self.HoldAStart.size
n_dirs = self.tasks.shape[0]
dirs_count = np.zeros(n_dirs, dtype=int)
# step over trials
for i in xrange(n_trials):
dir_idx = tc_util.get_task_idx(self.StartPos[i],
self.TargetPos[i],
self.tasks)
dirs_count[dir_idx] += 1
return dirs_count.max()
def make_binned(self, nbin=100,
align='speed',
verbose=False,
do_count=True,
do_rate=False):
'''Constructs aligned PSTHs of spikes in each direction.
Histograms are constructed with `n_bins`, aligned at increasing
and decreasing 15% speed points, windowed to include maximum
common period back to latest HoldAStart and up to earliest
HoldBFinish.
Parameters
----------
n_bins : int, default=100
number of bins
align : string
one of 'speed', 'hold', '3:2:3'
verbose : bool
print information messages?
Returns
-------
binned : BinnedData instance
bd.PSTHs.shape = (n_tasks, n_reps, n_dsets, nbins)
bd.pos.shape = (n_tasks, n_reps, nbins + 1, 3)
'''
#assert len(self.units) > 0
assert type(nbin) == int
assert align in align_types
align_starts, align_stops, bin_starts, bin_stops = \
self._get_limits(align)
do_spike = (do_count or do_rate)
# sort n_trials trials into n_dirs directions
# requires calc max_n_repeats per direction to construct array
max_repeat = self.calc_max_repeats()
ntrial = bin_starts.size
ndir = self.tasks.shape[0]
nunit = len(self.units)
if (nunit > 0) and do_spike:
if do_count:
counts = np.empty((ndir, max_repeat, nunit, nbin)) + np.nan
if do_rate:
rates = np.empty((ndir, max_repeat, nunit, nbin)) + np.nan
else:
rates = None
else:
counts = None
rates = None
bin_edges = np.empty((ndir, max_repeat, nbin + 1)) + np.nan
dirs_count = np.zeros(ndir)
pos = np.empty((ndir, max_repeat, nbin + 1, 3)) + np.nan
align_start_bins = None
align_stop_bins = None
for i in xrange(ntrial):
# directions are sorted according to the sorted unique tasks
if np.isnan(bin_starts[i]) or np.isnan(bin_stops[i]):
# couldn't align trial for some reason- ignore
if verbose:
print "Sort error, trial %d: " \
"couldn't find alignment points." \
% i
continue
# put index of this trial into correct direction column
dir_idx = tc_util.get_task_idx( \
self.StartPos[i], self.TargetPos[i], self.tasks)
bins = np.linspace(bin_starts[i], bin_stops[i], nbin + 1, \
endpoint=True)
if (nunit > 0) & do_spike:
if do_count:
trial_count = np.empty((nunit, nbin)) + np.nan
if do_rate:
trial_rate = np.empty((nunit, nbin)) + np.nan
for i_unit, unit in enumerate(self.units):
spikes = np.asarray(unit.spikes[i])
if do_count:
trial_count[i_unit], _ = np.histogram(spikes, bins=bins)
if do_rate:
trial_rate[i_unit], _ = tc_util.unbiased_histogram( \
np.asarray(unit.spikes[i]), bins=bins)
# interpolate position
these_pos = _interpolate_position(self.positions[i], bins)
# calculate (for display purposes later on) position of 15%-acc.
# in bin numbers - should be constant, but not integer, for each
bin_width = bins[1] - bins[0]
this_align_start = (align_starts[i] - bin_starts[i]) / bin_width
if align_start_bins != None:
assert np.allclose(this_align_start, align_start_bins)
align_start_bins = this_align_start
this_align_stop = (align_stops[i] - bin_starts[i]) / bin_width
if align_stop_bins != None:
assert np.allclose(this_align_stop, align_stop_bins)
align_end_bins = this_align_stop
# tests here
okay = True
max_pos_threshold = 0.1 # 10 cm
min_time_threshold = 0.05 # 50 ms
if np.any(np.abs(these_pos) > max_pos_threshold):
warn(AlignmentWarning("trial %d: position value too big" % i))
okay = False
if (align_stops[i] - align_starts[i]) < min_time_threshold:
warn(AlignmentWarning("trial %d: movement time too short" % i))
okay = False
# if passed tests, add to collection
if okay:
rep_idx = dirs_count[dir_idx]
if nunit > 0:
if do_count:
counts[dir_idx, rep_idx] = trial_count
if do_rate:
rates[dir_idx, rep_idx] = trial_rate
bin_edges[dir_idx, rep_idx] = bins
pos[dir_idx, rep_idx] = these_pos
dirs_count[dir_idx] += 1
#... keep track of next row to enter for this direction
unit_names = [u.unit_name for u in self.units]
lags = [u.lag for u in self.units]
return BinnedData(bin_edges, pos, self.tasks, unit_names, align,
lags=lags, count=counts, unbiased_rate=rates,
align_start_bins=align_start_bins,
align_end_bins=align_end_bins,
files=self.files)