def set_axis_layout(E, ax, bin_type, labelsize=8):
# if bin_type.endswith('bins'):
# # if bin_type == '12bins'
# major_xticks, minor_xticks = (4, 2)
# # elif bin_type == '24bins':
# # major_xticks, minor_xticks = (24, 1)
# else:
# stop
# formatter = '%d' if not E.BIN_SHIFT else '%1.1f'
# ax.xaxis.set_major_locator(MultipleLocator(major_xticks))
# ax.xaxis.set_major_formatter(FormatStrFormatter(formatter))
# ax.xaxis.set_minor_locator(MultipleLocator(minor_xticks))
xticks, xlabels = get_CT_ticks_labels(bin_type, SHIFT=E.BIN_SHIFT, tshift=E.binTimeShift)
rotation = 0
if E.BIN_SHIFT:
labelsize = 2
if bin_type.endswith('bins'):
rotation = 90
ax.set_xticks(xticks)
ax.set_xticklabels(xlabels, rotation=rotation)
ax.set_xlim([xticks[0]-.5, xticks[-1]+.5])
ax.set_ylim([0, ax.get_ylim()[1]])
ax.tick_params(axis='both', which='both', labelsize=labelsize, direction='out')
# DC grey background
ax.axvspan(12, 24, fc='.9', ec='.9', zorder=-1)
ax_cleanup(ax)
def set_panel_layout(E, fig, ax, feats, num_subplot, ymaxs, num_bins=12):
num_days = len(E.daysToUse)
ax.set_xlim([0, num_days * num_bins + 1])
if ymaxs is not None:
ax.set_ylim([0, ymaxs[num_subplot]])
xticks = np.arange(0, num_days * num_bins + 1, 6)
ax.set_xticks(xticks)
ax.set_xticklabels(xticks * 2)
ylims = np.array(ax.get_ylim())
if feats[num_subplot] == 'ASP':
ylims = [0, 1]
ax.set_ylim([0, ylims[1]])
for x in xticks[::2]:
ax.axvline(x=x, ymax=ylims[1], color='.7', zorder=0)
cnt = E.daysToUse[0]
for x in xticks[1::2]:
ax.text(x,
.9 * ylims[1],
'day%d' % cnt,
fontsize=6,
color='.7',
ha='center')
cnt += 1
ax.tick_params(axis='both', which='both', labelsize=7)
ax.set_aspect('auto')
ax_cleanup(ax)
plt.subplots_adjust(left=0.1, right=0.88, hspace=0.6) #, wspace=0.15)
def plot_features_distribution_binned_and_avgs_by_strain(E, bin_type='24HDCLC', NORMED=False):
features = E.features
_, _, _, _, md_data, md_labels = E.load_feature_vectors(features, bin_type)
if bin_type == '24HDCLC':
bins = ['24H', 'DC', 'LC']
elif bin_type == '12bins':
bins = ['CT%d-%d' %(x, x+2) for x in range(6, 26, 2)] + \
['CT%d-%d' %(x, x+2) for x in range(2, 6, 2)]
level = 'mouseday'
hist_bins = 50
for b, tbin in enumerate(bins[:1]):
dirname_out = os.path.join(E.figures_dir_subpar, 'features/vectors/distributions/%s/strain_breakdown/%s/' %(bin_type, tbin))
if not os.path.isdir(dirname_out): os.makedirs(dirname_out)
for strain in xrange(E.num_strains):
idx = md_labels[:, 0] == strain
data = md_data
fig, axes = plt.subplots(8, 3, figsize=(8, 12))
for f, feat in enumerate(features):
ax = axes.flatten()[f]
arr = data[f, idx, b]
# set xmax
x_max = np.mean(data[f]) + 5*np.std(data[f])
hist_bins = np.linspace(0, x_max, 50)
weights = np.ones_like(arr)
if NORMED:
weights /=len(arr)
color = E.plot_settings_dict[features[f]]['color']
y, _, _ = ax.hist(arr, bins=hist_bins, weights=weights, color=color)
ax.set_xlim(0, x_max)
ax.tick_params(axis='both', labelsize=6)
plt.subplots_adjust(left=0.1, right=0.88, hspace=0.6)
ax.set_title(feat, fontsize=10)
plotting_utils.ax_cleanup(ax)
fig.text(0.05, 0.95, E.strain_names[strain], fontsize=16)
fig.suptitle('Features distributions\nbinned and averages\nbinned%s %s' %(tbin, level))
fname = dirname_out + "%s_all_features_distribution_binavgs_%s_strain%d.pdf" %(level, tbin, strain)
fig.savefig(fname, bbox_inches='tight')
print "saved to:", fname
plt.close()
def plot_features_distribution_all_data(E, NORMED=False, GENERATE=False, VERBOSE=False):
"""24H and mouseday only
"""
features = E.features[:3]
# features dictionary each key (feature) is a list - 16 strains -
# of lists - all feature values during days -
all_data = {}
for feat in features:
all_data[feat] = E.generate_feature_arrays(feat, GENERATE, VERBOSE)
return all_data
# # manipulate the list
# arr_list = [list(x) for x in feat_arrays]
# data.append(np.array(list(itertools.chain(*arr_list))))
hist_bins = 50
# fig, ax = plt.subplots(figsize=(8, 12))
fig, axes = plt.subplots(8, 3, figsize=(8, 12))
for f, feat in enumerate(features):
ax = axes.flatten()[f]
arr = data[f]
# set xmax
x_max = np.mean(data[f]) + 5*np.std(data[f])
hist_bins = np.linspace(0, x_max, 50)
weights = np.ones_like(arr)
if NORMED:
weights /=len(arr)
color = E.plot_settings_dict[features[f]]['color']
y, _, _ = ax.hist(arr, bins=hist_bins, weights=weights, color=color)
ax.set_xlim(0, x_max)
ax.tick_params(axis='both', labelsize=6)
plt.subplots_adjust(left=0.1, right=0.88, hspace=0.6)
ax.set_title(feat, fontsize=10)
plotting_utils.ax_cleanup(ax)
fig.suptitle('Features distributions\nall data 24H from all mousedays')
dirname_out = os.path.join(E.figures_dir_subpar, 'features/vectors/distributions/')
if not os.path.isdir(dirname_out): os.makedirs(dirname_out)
fname = dirname_out + "mousedays_all_features_distribution_all_data_24H.pdf"
fig.savefig(fname, bbox_inches='tight')
print "saved to:", fname
plt.close()
return feat_arrays
def set_panel_layout(E, fig, ax, feats, num_subplot, ymaxs):
# xticks = E.daysToUse
ax.set_xlim([xticks[0] - 1, xticks[-1] + 1])
if ymaxs is not None:
ax.set_ylim([0, ymaxs[num_subplot]])
if len(E.daysToUse) > 20:
xticks = E.daysToUse[::2]
ax.set_xticks(xticks)
ax.set_xticklabels(xticks)
ylims = np.array(ax.get_ylim())
if feats[num_subplot] == 'ASP':
ylims = [0, 1]
ax.set_ylim([0, ylims[1]])
ax.tick_params(axis='both', which='both', labelsize=7)
ax.set_aspect('auto')
ax_cleanup(ax)
plt.subplots_adjust(left=0.1, right=0.88, hspace=0.6) #, wspace=0.15)
def plot_explained_variance_ratio(experiment, X, pca, level, lastcomp, fname,
EXCLUDE_SS):
E = experiment
components = pca.components_
num_components = pca.n_components_
expl_variance = pca.explained_variance_ratio_
figsize, nrows, ncols, _, _ = get_subplot_specs(E, num_subplots=2)
fig, (ax0, ax1) = plt.subplots(nrows=nrows, ncols=ncols, figsize=figsize)
ax0.plot(range(len(components)), expl_variance, color='.4')
ax1.plot(range(len(components))[-lastcomp:],
expl_variance[-lastcomp:],
color='.4')
ax0.set_ylim([0, .35])
ax1.set_ylim([0, .0006])
ax1.xaxis.set_major_locator(MultipleLocator(1))
for ax in [ax0, ax1]:
ax.xaxis.set_major_formatter(FormatStrFormatter('%d'))
ax_cleanup(ax)
# annotate close-up
text = 'close up,\nlast %d components' % lastcomp
ax1.text(.5, .7, text, fontsize=8, transform=ax1.transAxes)
text = '%d strains, ' %E.num_strains if not EXCLUDE_SS \
else '%d strains (no AKR, CAST, CZECH), ' %(E.num_strains-3)
level_ = 'mice' if level == 'mouse' else level
fig_title = 'Explained variance ratio, %d principal components\n%s%d %s, 24H avg data' % (
num_components, text, X.shape[0], level_)
subplot_labels(E,
fig,
xlabel='Principal component',
ylabel='explained variance',
fig_title=fig_title)
plt.subplots_adjust(bottom=.15, wspace=.4)
save_plot(fname, fig)
def plot_features_panel_LC_DC_ratio(experiment,
level='group',
bin_type='3cycles',
err_type='sem',
plot_type='features',
ADD_SOURCE=True):
E = experiment
print "%s, level: %s, bin_type: %s, err_type: %s, days: %s" % (
plot_features_panel_LC_DC_ratio.__name__, level, bin_type, err_type,
E.daysToUse)
features = E.features_by_activity
num_features = len(features)
days_to_use = E.daysToUse
dirname = E.figures_dir + '%s/vectors/panels_expdays/%s/%s_days/' % (
plot_type, bin_type, E.use_days)
if not os.path.isdir(dirname): os.makedirs(dirname)
fname = dirname + '%s_panel_expdays_LC_DC_ratio' % (plot_type
) #s, err_type)
# using mouseday value
# ratios = E.generate_feature_vectors_LC_DC_ratio_expdays(features, level, bin_type, err_type)
# using group averages value
data_ = E.generate_feature_vectors_expdays(features, level, bin_type,
err_type)
ratios_ = data_[:, :, :, 2, 0] / data_[:, :, :, 1, 0] # LC/DC, avgs
figsize = (14.4, 4.8)
nrows, ncols = 4, 7
fig, axes = plt.subplots(nrows, ncols, figsize=figsize, sharex=True)
xs = days_to_use
cnt = 0
for n in xrange(nrows):
for m in xrange(ncols):
ax = axes.flatten()[n * ncols + m]
if n == 0 and m >= 3:
ax.axis('off')
continue
feature = features[cnt]
colors = get_two_group_features_colors(E, feature)
subtitle = '%s %s' % (feature,
E.plot_settings_dict[feature]['unit'])
for g in xrange(E.num_strains):
# # using mouseday values
# tkw = {'fmt':'o-', 'ms':2, 'elinewidth':.5, 'capsize':1}
# ax.errorbar(xs, ratios[g, cnt, :, 0], yerr=ratios[g, cnt, :, 1],
# color=colors[g], mec=colors[g], **tkw)
# using group averages value
tkw = {'marker': 'o', 'ms': 2, 'zorder': 0}
ax.plot(xs,
ratios_[g, cnt, :],
color=colors[g],
mec=colors[g],
**tkw)
# use when plotting all for comparison
# colors_ = ['r', 'y']
# ax.plot(xs, ratios_[g, cnt, :], color=colors_[g], mec=colors_[g], **tkw)
if E.short_name == 'HFD2':
d_day = E.dietChangeDayNumbers[0]
ax.axvspan(xmin=d_day,
xmax=days_to_use[-1],
color='.9',
zorder=0)
xlims = ax.get_xlim()
ax.set_xlim((4, xlims[1]))
ax.set_ylim((0, ax.get_ylim()[1]))
ax_cleanup(ax)
ax.set_title(subtitle, fontsize=10)
cnt += 1
fig.text(.6,
.9,
'0: %s, control, 24H (grayscale)' % E.strain_names[0],
fontsize=8)
fig.text(.6,
.85,
'1: %s, knock-out, 24H (colors)' % E.strain_names[1],
fontsize=8)
fig.text(.6, .8, 'gray background: diet change', fontsize=8)
xlabel = 'Experiment Day'
add_xylabels(fig, xlabel=xlabel, xypad=-.02, labelsize=10)
days_string = E.get_days_to_use_text()
fig_title = '%s Experiment\nFeatures vs days, LC/DC ratio %s avg$\pm$%s\nacross %s days: %s' % (
E.short_name, level, err_type, E.use_days.replace(
'_', '-').title(), days_string)
add_titles_notes(E, fig, title=fig_title)
plt.subplots_adjust(hspace=.6, wspace=.34)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE, sypad=-.02)
def plot_features_panel_DC_LC(experiment,
level='group',
bin_type='3cycles',
err_type='sem',
plot_type='features',
ADD_SOURCE=True):
E = experiment
print "%s, level: %s, bin_type: %s, err_type: %s, days: %s" % (
plot_features_panel_DC_LC.__name__, level, bin_type, err_type,
E.daysToUse)
features = E.features_by_activity
num_features = len(features)
days_to_use = E.daysToUse
dirname = E.figures_dir + '%s/vectors/panels_expdays/%s/%s_days/' % (
plot_type, bin_type, E.use_days)
if not os.path.isdir(dirname): os.makedirs(dirname)
fname = dirname + '%s_panel_expdays_DC_LC_group_%s' % (plot_type, err_type)
# groups, features, days, cycles, avg/err
data = E.generate_feature_vectors_expdays(features, level, bin_type,
err_type)
new_data = data[:, :, :, 1:, :]
figsize = (14.4, 4.8)
nrows, ncols = 4, 7
fig, axes = plt.subplots(nrows, ncols, figsize=figsize, sharex=True)
xs = days_to_use
cnt = 0
for n in xrange(nrows):
for m in xrange(ncols):
ax = axes.flatten()[n * ncols + m]
if n == 0 and m >= 3:
ax.axis('off')
continue
feature = features[cnt]
# get colors
if feature.startswith('AS'):
color1, color2 = [E.fcolors['AS'].values()[x] for x in [1, 3]]
elif feature.startswith('T'):
color1, color2 = [
E.fcolors[feature[1:]].values()[x] for x in [1, 3]
]
else:
color1, color2 = [
E.fcolors[feature[:1]].values()[x] for x in [1, 3]
]
colors = ['.4', '.7', color1, color2]
subtitle = '%s %s' % (feature,
E.plot_settings_dict[feature]['unit'])
for g in xrange(E.num_strains):
tkw = {'fmt': 'o-', 'ms': 2, 'elinewidth': .5, 'capsize': 1}
# DC
mec = '0' if g == 0 else colors[2 * g]
ax.errorbar(xs,
new_data[g, cnt, :, 0, 0],
yerr=new_data[g, cnt, :, 0, 1],
color=colors[2 * g],
mec=mec,
**tkw)
#LC
mc = '0' if g == 0 else colors[2 * g + 1]
ax.errorbar(xs,
new_data[g, cnt, :, 1, 0],
yerr=new_data[g, cnt, :, 1, 1],
color=colors[2 * g + 1],
mec=mec,
**tkw)
if E.short_name == 'HFD2':
d_day = E.dietChangeDayNumbers[0]
ax.axvspan(xmin=d_day,
xmax=days_to_use[-1],
color='.9',
zorder=0)
xlims = ax.get_xlim()
ax.set_xlim((4, xlims[1]))
ax.set_ylim([0, ax.get_ylim()[1]])
ax_cleanup(ax)
ax.set_title(subtitle, fontsize=10)
cnt += 1
fig.text(.6,
.9,
'0: %s, control (grayscale)\nDC:black, LC: gray' %
E.strain_names[0],
fontsize=8)
fig.text(.6,
.83,
'1: %s, knock-out (colors)\nDC: dark, LC: light ' %
E.strain_names[1],
fontsize=8)
fig.text(.6, .78, 'gray background: diet change', fontsize=8)
xlabel = 'Experiment Day'
add_xylabels(fig, xlabel=xlabel, xypad=-.02, labelsize=10)
days_string = E.get_days_to_use_text()
fig_title = '%s Experiment\nFeatures, DC vs. LC, %s avg and %s\nacross %s days: %s' % (
E.short_name, level, err_type, E.use_days.replace(
'_', '-').title(), days_string)
add_titles_notes(E, fig, title=fig_title)
plt.subplots_adjust(hspace=.6, wspace=.34)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE, sypad=-.02)
def plot_features_panel_12bins(experiment,
level='group',
bin_type='12bins',
err_type='sem',
plot_type='features',
ADD_SOURCE=True):
E = experiment
print "%s, level: %s, bin_type: %s, err_type: %s, days: %s" % (
plot_features_panel_12bins.__name__, level, bin_type, err_type,
E.daysToUse)
features = E.features_by_activity
num_features = len(features)
days_to_use = E.daysToUse
num_bins = int(bin_type.strip('bins'))
tbins = my_utils.get_tbins_string(bin_type)
dirname = E.figures_dir + '%s/vectors/panels_expdays/%s/%s_days/' % (
plot_type, bin_type, E.use_days)
if not os.path.isdir(dirname): os.makedirs(dirname)
data = E.generate_feature_vectors_expdays(features, level, bin_type,
err_type)
ymaxs = get_ymaxs(data)
# stop
for b in xrange(num_bins):
new_data = data[:, :, :, b, :]
figsize = (14.4, 4.8)
nrows, ncols = 4, 7
fig, axes = plt.subplots(nrows, ncols, figsize=figsize, sharex=True)
xs = days_to_use
cnt = 0
for n in xrange(nrows):
for m in xrange(ncols):
ax = axes.flatten()[n * ncols + m]
if n == 0 and m >= 3:
ax.axis('off')
continue
feature = features[cnt]
colors = get_two_group_features_colors(E, feature)
subtitle = '%s %s' % (feature,
E.plot_settings_dict[feature]['unit'])
for g in xrange(E.num_strains):
tkw = {
'fmt': 'o-',
'ms': 2,
'elinewidth': .5,
'capsize': 1
}
ax.errorbar(xs,
new_data[g, cnt, :, 0],
yerr=new_data[g, cnt, :, 1],
color=colors[g],
mec=colors[g],
**tkw)
if E.short_name == 'HFD2':
d_day = E.dietChangeDayNumbers[0]
ax.axvspan(xmin=d_day,
xmax=days_to_use[-1],
color='.9',
zorder=0)
# xmin, xmax = ax.get_xlim()
# ax.set_xlim((xmin-.5, xmax+.5))
# ax.set_ylim([0, ymaxs[cnt]])
ax_cleanup(ax)
ax.set_title(subtitle, fontsize=10)
cnt += 1
xmin, xmax = ax.get_xlim()
ax.set_xlim((xmin - .5, xmax + .5))
fig.text(.6,
.9,
'0: %s, control, %s (grayscale)' %
(E.strain_names[0], tbins[b]),
fontsize=8)
fig.text(.6,
.85,
'1: %s, knock-out, %s (colors)' %
(E.strain_names[1], tbins[b]),
fontsize=8)
fig.text(.6, .8, 'gray background: diet change', fontsize=8)
xlabel = 'Experiment Day'
add_xylabels(fig, xlabel=xlabel, xypad=-.02, labelsize=10)
days_string = E.get_days_to_use_text()
fig_title = '%s Experiment\nFeatures, %s %s avg$\pm$%s, %s\nacross %s days: %s' % (
E.short_name, bin_type, level, err_type, tbins[b],
E.use_days.replace('_', '-').title(), days_string)
add_titles_notes(E, fig, title=fig_title)
plt.subplots_adjust(hspace=.6, wspace=.34)
fname = dirname + '%s_panel_expdays_%s_group_%s_bin%d' % (
plot_type, bin_type, err_type, b)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE, sypad=-.02)
def plot_HCM_system_time(E, ttype='maintenance', ADD_SOURCE=True):
"""plots system_on_off
"""
if ttype == 'maintenance':
HCM_time, labels = E.get_HCM_maintenance_time()
xlims = [2, 8]
xticks = range(2, 10+1, 2)
xticklabels = xticks
elif ttype == 'recording':
HCM_time, labels = E.get_HCM_recording_time()
xlims = [2, 34]
xticks = range(4, 32+1, 4)
xticklabels = [str(x) for x in range(4, 24+1, 4)] \
+ [str(x) for x in range(4, 10+1, 4)]
print "%s, %s days.." %(plot_HCM_system_time.__name__, E.use_days)
figsize, nrows, ncols, sharex, sharey = \
get_subplot_specs(E, num_subplots=len(E.groups), plot_type='subplot')
fig, axes = plt.subplots(nrows, ncols, figsize=figsize, sharex=sharex, sharey=sharey)
for strain in range(nrows * ncols):
ax = axes.flatten()[strain] if type(axes) == np.ndarray else axes
idx = labels [:, 0] == strain
data = HCM_time[idx] / 3600. - 7
c = 0
for row in data:
ax.plot([row[0], row[1]], [c, c], lw=0.5, color='0.5')
c += 1
ax.set_xlim(xlims)
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels)
if ttype == 'maintenance':
text = 'avg:\n%1.1fmin' %(np.diff(data).mean() * 60)
elif ttype == 'recording':
text = 'avg:\n%1.2fhrs' %np.diff(data).mean()
ax.axvline(x=12, linestyle='--', color='k')
ax.axvline(x=24, linestyle='--', color='k')
ax.text(.9, .9, text, fontsize=10,
ha='center', transform=ax.transAxes)
ax.set_xlabel('Circadian Time')
ax.set_ylabel('Obs. days', labelpad=10)
ax.set_title('group%d: %s' %(strain, E.strain_names[strain]), y=1.05)
ax_cleanup(ax, labelsize=10)
title = '%sExp\n%s days\nDaily %s Time' %(
E.short_name, E.use_days.replace('_', '-').title(), ttype.title())
add_titles_notes(E, fig,
title=title,
typad=.18,
tlxpad=-.1,
tlypad=.15,
HP_TEXT=False)
plt.subplots_adjust(wspace=0.4)
# save
dirname_out = E.figures_dir + 'HCM_recording_time/%s_days/' %E.use_days
if not os.path.isdir(dirname_out): os.makedirs(dirname_out)
fname = dirname_out + 'HCMsystem_%s_time_%s_days' %(ttype, E.use_days)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE, sypad=-.1)
def plot_loadings(experiment,
X,
pca_components,
feature_names,
level,
comps_idx,
fname,
plot_type,
EXCLUDE_SS,
ADD_SOURCE=True):
E = experiment
num_items = X.shape[0]
num_components = len(pca_components)
num_features = len(feature_names)
assert num_features == num_components, 'wtf'
# select components
selected_comps = pca_components[comps_idx]
fig, ax = plt.subplots(figsize=(7, 3))
bar_width = .03 if len(comps_idx) == num_components else .1
colors = [E.plot_settings_dict[feat]['color'] for feat in feature_names]
if plot_type == 'bar':
for c, comp in enumerate(selected_comps):
ind = np.arange(num_features) + c * bar_width
alpha_step = .03 * comp if len(
comps_idx) == num_components else .2 * comp
tkw = {
'lw': .2,
'alpha': (1 - alpha_step),
'color': colors,
'edgecolor': None,
'label': 'PC%d' % comp
}
ax.bar(ind, pca_components[-(comp + 1)], bar_width, **tkw)
elif plot_type == 'line':
colors = get_16_colors()[1::4]
tkw = {
'numpoints': 1,
'fontsize': 6,
'handletextpad': 1.2,
'labelspacing': .8,
'borderaxespad': 1.5,
'frameon': False,
'bbox_to_anchor': (1., 1.2),
'bbox_transform': ax.transAxes
}
for c, comp in enumerate(selected_comps):
ax.plot(comp, color=colors[c], label='comp%d' % (comps_idx[c]))
leg = ax.legend(**tkw)
ax.set_xticks(range(num_features))
ax.set_xticklabels([])
ax.set_ylim(-1, 1)
ax.axhline(xmin=0, xmax=num_features, lw=.5, color='k')
ax.spines['bottom'].set_position('zero')
ax.spines['bottom'].set_visible(False)
ax_cleanup(ax)
# display feature names on top of bars
xpos = np.arange(num_features) + .2
heights = selected_comps.max(axis=0) + .04
heights[5] = heights[4] + .05 # correct FBASR
if plot_type == 'line': heights += .1
for p, feat in enumerate(feature_names):
ax.text(xpos[p],
heights[p],
feat,
fontsize=6,
ha='center',
transform=ax.transData)
# annotate
if plot_type == 'hist':
annotation = 'Color-code by feature type.\n' \
+ 'AS:purple, Feeding:orange, Drinking:blue, Move:green.\n' \
+ 'For each feature, components from %d (opaque) to\n' %comps_idx[0] \
+ '%dth (transparent) are shown on a sliding\ntransparency scale.' %comps_idx[-1]
ax.text(13.8,
ax.get_ylim()[0],
annotation,
fontsize=6,
va='center',
transform=ax.transData)
# title
text = '%d strains, ' %E.num_strains if not EXCLUDE_SS \
else '%d strains (no AKR, CAST, CZECH), ' %(E.num_strains-3)
level_ = 'mice' if level == 'mouse' else level
fig_title = 'Principal Component loadings, components: %s of %d total\n%s%d %s, 24H avg data\n%s days:%dto%d' % (
comps_idx, num_components, text, num_items, level_, E.use_days,
E.daysToUse[0], E.daysToUse[-1])
subplot_labels(E, fig, xlabel='', ylabel='Loading', fig_title=fig_title)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE)
def plot_component_loadings(E,
X,
pca_components,
feature_names,
level,
fname,
EXCLUDE_SS=False,
BY_STRAIN=False,
strain=None,
ADD_SOURCE=True):
num_items = X.shape[0]
num_components = len(pca_components)
num_features = len(feature_names)
figsize, nrows, ncols, sharex, sharey = get_subplot_specs(
E, num_subplots=num_components)
fig, axes = plt.subplots(nrows,
ncols,
figsize=figsize,
sharex=sharex,
sharey=sharey)
xs = range(1, num_components + 1)
for c, comp in enumerate(pca_components):
ax = axes.flatten()[c]
ax.plot(xs, comp, color='.4')
ax.set_xlim(0, ax.get_xlim()[1])
ax.set_xticks(range(num_features))
ax.set_xticklabels([])
ax.set_title('PC%d' % c, fontsize=8, position=(.5, 1.15), ha='center')
ax.axhline(xmin=0, xmax=num_features, lw=.5, color='k')
ax.spines['bottom'].set_position('zero')
ax.spines['bottom'].set_visible(False)
ax_cleanup(ax)
# display feature names on top
heights = list(np.linspace(1.05, .55, 4)) * 8
for p, feat in enumerate(feature_names):
ax.text(xs[p],
heights[p],
feat,
fontsize=4,
ha='center',
transform=ax.transData)
ax.axvline(x=xs[p], ymin=.4, ymax=heights[p], lw=.2, color='.5')
# title
text = '%d strains,' %E.num_strains if not EXCLUDE_SS \
else '%d strains (no AKR, CAST, CZECH),' %(E.num_strains-3)
level_ = 'mice' if level == 'mouse' else level
fig_title = 'Principal Component loadings, %d components\n' %num_components \
+ '%s %d %s, 24H avg data\n%s days:%dto%d' %(
text, num_items, level_, E.use_days, E.daysToUse[0], E.daysToUse[-1])
if BY_STRAIN:
fig_title = 'Principal Component loadings, %d components\n' %num_components \
+ '%s %d %s, 24H avg data\ngroup%d: %s\n%s days:%dto%d' %(
text, num_items, level_, strain, E.strain_names[strain],
E.use_days, E.daysToUse[0], E.daysToUse[-1])
subplot_labels(E, fig, xlabel='', ylabel='Loading', fig_title=fig_title)
plt.subplots_adjust(hspace=.4, wspace=.3)
save_plot(fname, fig, ADD_SOURCE=ADD_SOURCE)
