def generate_position_bin_times(self, cycle=None, tbin_type=None, xbins=12,
ybins=24, GENERATE=False):
""" returns total time in cage divided into xbins, ybins
"""
name = 'bin_times_xbins%d_ybins%d' %(xbins, ybins)
text = my_utils.get_text(cycle, tbin_type)
varName = 'bin_times_%s_xbins%d_ybins%d' %(text, xbins, ybins)
dirname = self.experiment.derived_dir + \
'position_data/xbins%d_ybins%d/%s/bin_times/' %(
xbins, ybins, text)
if not os.path.isdir(dirname): os.makedirs(dirname)
if GENERATE:
bin_times = self.get_position_bin_times(cycle, tbin_type, xbins, ybins)
np.save(dirname + self.id_string, bin_times)
else:
try:
bin_times = self.load(varName)
# print bin_times
except IOError:
bin_times = self.get_position_bin_times(cycle, tbin_type, xbins, ybins)
np.save(dirname + self.id_string, bin_times)
setattr(self, varName, bin_times)
return bin_times
def check_position_bin_times(self, xbins=12, ybins=24, tbin_type=None):
stop
text = my_utils.get_text(tbin_type=tbin_type)
if tbin_type.endswith('bins'):
data = self.generate_position_bin_times(tbin_type=tbin_type, xbins=xbins, ybins=ybins)
if not AS_FLAG:
data_ = self.generate_position_bin_times(
cycle='24H', xbins=xbins, ybins=ybins)
# tolerance: /num_seconds, /1000 tolerates the 1000secs~=10mins
np.testing.assert_array_almost_equal(data.sum(0)/10000., data_/10000.,
decimal=0, err_msg='24H != sum over %s' %tbin_type)
else:
data_ = self.generate_position_bin_times(
cycle='AS24H', xbins=xbins, ybins=ybins)
# tolerance: /num_seconds, /1000 tolerates the 1000secs~=10mins
np.testing.assert_array_almost_equal(data.sum(0)/1000., data_/1000.,
decimal=0, err_msg='AS24H != sum over %s' %tbin_type)
else:
cycles = self.experiment.all_cycles
data = np.zeros((len(cycles), ybins, xbins))
c = 0
for cycle in cycles:
data[c] = self.generate_position_bin_times(cycle=cycle, xbins=xbins, ybins=ybins)
c +=1
h24, IS, dc, lc, as24, asdc, aslc = data
# tolerance: /num_seconds, /1000 tolerates the 1000secs~=10mins
np.testing.assert_array_almost_equal(h24/10000., (dc+lc)/10000., decimal=0,
err_msg='24H != DC+LC')
np.testing.assert_array_almost_equal(h24, IS+as24, decimal=0,
err_msg='24H != IS+AS')
np.testing.assert_array_almost_equal(as24/1000., (asdc+aslc)/1000., decimal=0,
err_msg='AS24 != ASDC+ASLC')
print "%s : POSITION DENSITY CHECKS, %s, xbins:%s, ybins:%d -> OK" %(
self, text, xbins, ybins)
def get_figure_titles(E,
labels,
level,
act=None,
cycle=None,
tbin_type=None,
err_type=None,
plot_type=None):
days_string = E.get_days_to_use_text()
text = my_utils.get_text(cycle, tbin_type)
# labels are a bunch of labels
fig_title = '%s Experiment\n%s, %s, %s\n%s days: %s' % (
E.short_name,
plot_type.replace('_', ' ').title(),
level,
text,
E.use_days.replace('_', ' ').upper(),
days_string,
)
if level == 'group':
subtitles = [
'%d: %s' % (n, name) for n, name in zip(labels, E.strain_names)
]
elif level == 'mouse':
strain, mouseNumbers = np.unique(labels[:, 0])[0], labels[:, 1]
subtitles = ['M%d' % n for n in mouseNumbers]
fig_title += '\ngroup%d: %s' % (strain, E.strain_names[strain])
elif level == 'mouseday':
strain, mouseNumber, days = np.unique(labels[:, 0])[0], np.unique(
labels[:, 1])[0], labels[:, 2]
subtitles = ['D%d' % n for n in days]
fig_title += '\ngroup%d: %s, M%d' % (strain, E.strain_names[strain],
mouseNumber)
return fig_title, subtitles
def generate_position_density(self,
cycle='24H',
tbin_type=None,
level='mouseday',
xbins=12,
ybins=24,
days=None,
err_type='sd',
GENERATE=False,
VERBOSE=False):
""" returns position density data
"""
cstart = time.clock()
if days is None:
days_to_use = self.daysToUse
num_md = self.num_mousedays
else:
days_to_use = days
num_md, _ = self.count_mousedays(days=days_to_use)
if tbin_type is not None:
num_tbins = my_utils.get_CT_bins(
tbin_type=tbin_type.strip('AS')).shape[0]
md_data = np.zeros((num_md, num_tbins, ybins, xbins))
else:
md_data = np.zeros((num_md, ybins, xbins))
text = my_utils.get_text(cycle, tbin_type)
print "Exp: %s, %s, tbin_type: %s, level:%s, xbins=%d, ybins=%d, err_type: %s, %s days" % (
self.short_name, self.generate_position_density.__name__, text,
level, xbins, ybins, err_type, days_to_use)
md_labels = np.zeros((num_md, 3), dtype=int)
cnt = 0
for group in self.groups:
for mouse in group.individuals:
for MD in mouse.mouse_days:
if MD.dayNumber in days_to_use:
if MD.ignored:
md_data[cnt] = np.nan
else:
bin_times = MD.generate_position_bin_times(
cycle, tbin_type, xbins, ybins, GENERATE)
if tbin_type is not None:
b = 0
for btimes in bin_times:
if btimes.sum() >= 0:
md_data[cnt, b] = btimes / btimes.sum()
b += 1
else:
if bin_times.sum() >= 0:
md_data[cnt] = bin_times / bin_times.sum()
md_labels[
cnt] = MD.groupNumber, MD.mouseNumber, MD.dayNumber
# if GENERATE and not VERBOSE:
# my_utils.print_progress(cnt, self.num_md_ok)
cnt += 1
cstop = time.clock()
if GENERATE:
if not VERBOSE:
print
dirname = self.derived_dir + \
'position_data/xbins%d_ybins%d/%s/bin_times/' %(xbins, ybins, text) # copied from: MD.generate_position_bin_times
print "binary output saved to: %s" % dirname
print "..took %1.2f minutes" % ((cstop - cstart) / 60.)
E = self
data, labels = my_utils.get_2d_averages_errors(E, md_data, md_labels,
tbin_type, level,
err_type)
return data, labels
def write_to_csv(experiment,
cycle='24H',
tbin_type=None,
AS_FLAG=False,
level='group',
xbins=12,
ybins=24,
err_type='sd',
plot_type='position_density'):
E = experiment
use_days = E.use_days + '_days'
text = my_utils.get_text(cycle, tbin_type, AS_FLAG)
print "%s, %s, %s, xbins:%d, ybins:%d, %s" % (
write_to_csv.__name__, text, level, xbins, ybins, use_days)
suffix = '%s/csv_files/xbins%d_ybins%d/%s/%s/%s_days/' % (
plot_type, xbins, ybins, level, text, E.use_days)
dirname = os.path.join(E.figures_dir, suffix)
if not os.path.isdir(dirname): os.makedirs(dirname)
fname = dirname + '%s_%s_%s_xbins%d_ybins%d.csv' % (plot_type, text, level,
xbins, ybins)
# load data
data, labels = E.generate_position_density(cycle=cycle,
tbin_type=tbin_type,
AS_FLAG=AS_FLAG,
level=level,
xbins=xbins,
ybins=ybins,
err_type=err_type)
if tbin_type is not None:
arr = my_utils.get_CT_bins(bin_type=tbin_type) / 3600 - 7
tbins = [
'bin%d: CT%d-%d' % (b, row[0], row[1]) for b, row in enumerate(arr)
]
new_data = data[:, :, :, :,
0].reshape(data.shape[0], data.shape[1],
-1) # (num_mousedays, 4x2 raveled)
else:
new_data = data[:, :, :, 0].reshape(data.shape[0],
-1) # (num_mousedays, 4x2 raveled)
# headers
# string = ', BIN_SHIFTED: %dm' %(E.binTimeShift/60) if E.BIN_SHIFT else ''
headers = [['Experiment: %s' % E.short_name],
['%s, xbins=%d, ybins=%d' % (plot_type, xbins, ybins)],
['%s, %s' % (level, text)], ['%s days' % E.use_days],
['days: %s' % E.daysToUse]]
if level == 'group':
sub_header = ['use_days', 'group']
if tbin_type is not None:
if (xbins, ybins) == (2, 4):
sub_header += [
'tbins', 'Niche', 'TopLeft', '[]', '[]', '[]', '[]',
'Feeder', 'Lick'
]
else:
sub_header += ['tbins']
else:
sub_header += [
'Niche', 'TopLeft', '[]', '[]', '[]', '[]', 'Feeder', 'Lick'
]
with open(fname, 'wb') as csv_file:
writer = csv.writer(csv_file, delimiter=",")
for h in headers:
writer.writerow(h)
writer.writerow('\n')
writer.writerow(sub_header)
writer.writerow('\n')
if tbin_type is not None:
new_data_ = new_data.swapaxes(
0, 1) # tbin, days, (raveled xbins, ybins)
for b in xrange(len(new_data_)):
bin_data = new_data_[b]
c = 0
for row in bin_data:
row_text = [
use_days + ', %s' % text, labels[c], tbins[b]
]
writer.writerow(
np.hstack([row_text, ['%1.5f' % r for r in row]]))
c += 1
writer.writerow('\n')
else:
c = 0
for row in new_data:
row_text = [use_days + ', %s' % text, labels[c]]
writer.writerow(
np.hstack([row_text, ['%1.5f' % r for r in row]]))
c += 1
writer.writerow('\n')
elif level == 'mouse':
sub_header = ['use_days', 'group', 'mouse']
if tbin_type is not None:
if (xbins, ybins) == (2, 4):
sub_header += [
'tbins', 'Niche', 'TopRight', '[]', '[]', '[]', '[]',
'Feeder', 'Lick'
]
else:
sub_header += ['tbins']
else:
sub_header += [
'Niche', 'TopLeft', '[]', '[]', '[]', '[]', 'Feeder', 'Lick'
]
with open(fname, 'wb') as csv_file:
writer = csv.writer(csv_file, delimiter=",")
for h in headers:
writer.writerow(h)
writer.writerow('\n')
writer.writerow(sub_header)
writer.writerow('\n')
for group_num in range(E.num_strains):
idx_group = labels[:, 0] == group_num
group_data = new_data[idx_group]
group_labels = labels[idx_group]
if tbin_type is not None:
group_data_ = group_data.swapaxes(
0, 1) # tbin, days, (raveled xbins, ybins)
for b in xrange(len(group_data_)):
bin_data = group_data_[b]
c = 0
for row in bin_data:
row_text = [
use_days + ', %s' % text, group_labels[c, 0],
group_labels[c, 1], tbins[b]
]
writer.writerow(
np.hstack(
[row_text, ['%1.5f' % r for r in row]]))
c += 1
# stop
writer.writerow('\n')
writer.writerow(['', '', '', 'Mean:'] + [
'%1.5f' % r for r in np.nanmean(bin_data, axis=0)
])
writer.writerow(
['', '', '', 'sd:'] +
['%1.5f' % r for r in np.nanstd(bin_data, axis=0)])
sem = np.nanstd(
bin_data, axis=0) / np.sqrt(bin_data.shape[0] - 1)
writer.writerow(['', '', '', 'sem:'] +
['%1.5f' % r for r in sem])
writer.writerow('\n')
else:
c = 0
for row in group_data:
row_text = [
use_days + ', %s' % text, group_labels[c, 0],
group_labels[c, 1]
]
writer.writerow(
np.hstack([row_text, ['%1.5f' % r for r in row]]))
c += 1
writer.writerow('\n')
writer.writerow(
['', '', 'Mean:'] +
['%1.5f' % r for r in np.nanmean(group_data, axis=0)])
writer.writerow(
['', '', 'sd:'] +
['%1.5f' % r for r in np.nanstd(group_data, axis=0)])
sem = np.nanstd(group_data,
axis=0) / np.sqrt(group_data.shape[0] - 1)
writer.writerow(['', '', 'sem:'] +
['%1.5f' % r for r in sem])
writer.writerow('\n')
elif level == 'mouseday':
sub_header = ['', 'group', 'mouse', 'mouseday']
if tbin_type is not None:
if (xbins, ybins) == (2, 4):
sub_header += [
'tbins', 'Niche', 'TopLeft', '[]', '[]', '[]', '[]',
'Feeder', 'Lick'
]
else:
sub_header += ['tbins']
else:
sub_header += [
'Niche', 'TopLeft', '[]', '[]', '[]', '[]', 'Feeder', 'Lick'
]
with open(fname, 'wb') as csv_file:
writer = csv.writer(csv_file, delimiter=",")
for h in headers:
writer.writerow(h)
writer.writerow('\n')
writer.writerow(sub_header)
writer.writerow('\n')
for group_num in range(E.num_strains):
idx_group = labels[:, 0] == group_num
group_data = new_data[idx_group]
group_labels = labels[idx_group]
m_labels = np.unique(group_labels[:, 1])
for mouse in m_labels:
idx_mouse = group_labels[:, 1] == mouse
m_data = group_data[idx_mouse]
m_labels = group_labels[idx_mouse]
if tbin_type is not None:
m_data_ = m_data.swapaxes(
0, 1) # tbin, days, (raveled xbins, ybins)
for b in xrange(len(m_data_)):
bin_data = m_data_[b]
c = 0
for row in bin_data:
row_text = [
use_days + ', %s' % text, m_labels[c, 0],
m_labels[c, 1], m_labels[c, 2], tbins[b]
]
writer.writerow(
np.hstack(
[row_text,
['%1.5f' % r for r in row]]))
c += 1
# stop
writer.writerow('\n')
writer.writerow(['', '', '', '', 'Mean:'] + [
'%1.5f' % r
for r in np.nanmean(bin_data, axis=0)
])
writer.writerow(['', '', '', '', 'sd:'] + [
'%1.5f' % r
for r in np.nanstd(bin_data, axis=0)
])
sem = np.nanstd(
bin_data,
axis=0) / np.sqrt(bin_data.shape[0] - 1)
writer.writerow(['', '', '', '', 'sem:'] +
['%1.5f' % r for r in sem])
writer.writerow('\n')
else:
c = 0
for row in m_data:
row_text = [
use_days + ', %s' % text, m_labels[c, 0],
m_labels[c, 1], m_labels[c, 2]
]
writer.writerow(
np.hstack(
[row_text, ['%1.5f' % r for r in row]]))
c += 1
writer.writerow('\n')
writer.writerow(
['', '', '', 'Mean:'] +
['%1.5f' % r for r in np.nanmean(m_data, axis=0)])
writer.writerow(
['', '', '', 'sd:'] +
['%1.5f' % r for r in np.nanstd(m_data, axis=0)])
sem = np.nanstd(m_data,
axis=0) / np.sqrt(m_data.shape[0] - 1)
writer.writerow(['', '', '', 'sem:'] +
['%1.5f' % r for r in sem])
writer.writerow('\n')
print "csv files saved to:\n%s" % fname
def plot_position_density(experiment,
cycle='24H',
tbin_type=None,
level='mouseday',
xbins=12,
ybins=24,
err_type='sd',
plot_type='position_density',
ADD_SOURCE=True):
E = experiment
text = my_utils.get_text(cycle, tbin_type)
print "%s, %s, %s, xbins:%d, ybins:%d, %s" % (
plot_position_density.__name__, text, level, xbins, ybins, E.use_days)
dirname_out = E.figures_dir + '%s/xbins%d_ybins%d/%s/%s/%s_days/' % (
plot_type, xbins, ybins, level, text, E.use_days)
if not os.path.isdir(dirname_out): os.makedirs(dirname_out)
# load data
data, labels = E.generate_position_density(cycle=cycle,
tbin_type=tbin_type,
level=level,
xbins=xbins,
ybins=ybins,
err_type=err_type)
if level == 'group':
fname = dirname_out + '%s_%s_%dgroups_xbins%d_ybins%d' % (
plot_type, text, E.num_strains, xbins, ybins)
draw_position_density(E,
data,
labels,
cycle,
tbin_type,
level,
err_type,
xbins,
ybins,
fname,
plot_type=plot_type,
ADD_SOURCE=ADD_SOURCE)
elif level == 'mouse':
for strain in xrange(E.num_strains):
idx = labels[:, 0] == strain
m_data, m_labels = data[idx], labels[idx]
fname = dirname_out + '%s_%s_group%d_xbins%d_ybins%d' % (
plot_type, text, strain, xbins, ybins)
draw_position_density(E,
m_data,
m_labels,
cycle,
tbin_type,
level,
err_type,
xbins,
ybins,
fname,
plot_type=plot_type,
ADD_SOURCE=ADD_SOURCE)
elif level == 'mouseday':
# nest
rects_HB = np.array([None] * len(data))
obs_rects = np.array([None] * len(data))
condition = (xbins, ybins) == (2, 4)
if condition:
rects_HB, obs_rects = E.generate_homebase_data()
for strain in range(E.num_strains):
idx_m = labels[:, 0] == strain
mouseNumbers = np.unique(labels[idx_m, 1])
for c in xrange(len(mouseNumbers)):
idx_md = labels[:, 1] == mouseNumbers[c]
md_data, md_labels = data[idx_md], labels[idx_md]
fname = dirname_out + '%s_%s_group%d_M%d_xbins%d_ybins%d' % (
plot_type, text, strain, mouseNumbers[c], xbins, ybins)
rect_HB, obs_rect = None, None
if condition:
idx_start, idx_end = my_utils.find_nonzero_runs(idx_md)[0]
rect_HB = rects_HB[idx_start:idx_end]
obs_rect = obs_rects[idx_start:idx_end]
draw_position_density(E,
md_data,
md_labels,
cycle,
tbin_type,
level,
err_type,
xbins,
ybins,
fname,
rect_HB=rect_HB,
obs_rect=obs_rect,
plot_type=plot_type,
ADD_SOURCE=ADD_SOURCE)