x = np.linspace(-r, r, num=200)
def f(x, a, mu, sig):
y = a * np.exp((-(x - mu)**2) / (2 * sig**2))
# y = a*np.exp(-np.power((x-mu), 2)/(2 * sig**2))
return y
y1 = f(x, .3, 0, 3)
y2 = f(x, -.1, 0, 9)
fig = texplot.figsize(.9)
plf.addarrowaxis(plt.gca(),
x=0.1,
y=0.4,
dx=0.1,
dy=0.2,
xtext='Distance',
ytext='Sensitivity',
xtextoffset=.02,
ytextoffset=.032,
fontsize='x-small')
plt.plot(x, y1 + y2)
plt.plot(x, y1, '--')
plt.plot(x, y2, '--')
plt.xticks([])
plt.yticks([])
texplot.savefig('rfstructure')
plt.show()
#plt.axis('equal')
plf.subplottext('C', ax3, x=-0.2, y=1.025)
ax3.set_xticks([0, 1])
ax3.set_xticklabels(['Mesopic', 'Photopic'])
ax3.set_ylabel('Polarity Index')
plf.spineless(ax3)
distrib = [len(group) * 100 / cells.shape[0] for group in groups]
ax4 = plt.subplot(2, 2, 4)
# Hack for fixing the whitespace on both sides for this subplot
# On (2,2) grid, there is too much space on the left of the fourth
# plot for some reason.
# Also , extra space is needed for the cell category names, setting the
# colspan to 4 instead of 5 gives enough space for them.
ax4 = plt.subplot2grid((2, 10), (1, 5), colspan=4)
ax4.barh(np.arange(5), distrib, tick_label=colorlabels, color=colorcategories)
# Labels look unaligned on ipython output but the saved pdf and pgf are fine
for txt in ax4.get_yticklabels():
txt.set_x(1.2)
txt.set_ha('right')
plf.spineless(ax4)
ax4.tick_params(axis='y', length=0)
ax4.invert_yaxis()
ax4.set_xlabel('Percentage of cells')
plf.subplottext('D', ax4, x=-0, y=1.15)
plt.subplots_adjust(hspace=.4, wspace=.4)
texplot.savefig('polarityindexchange')
plt.show()
groups = []
bins = np.arange(-.2, .35+0.05, 0.05)
for i, color in enumerate(colorcategories):
group = [index for index, c in enumerate(colors) if c == color]
ax = axes[i]
change = csichange[group]
groups.append(group)
changes.append(change)
if i == 0:
csichange_on = np.copy(change)
elif i == 1:
csichange_off = np.copy(change)
ax.hist(change, color=color , bins=bins)
ax.set_ylim([0, 10])
ax.set_yticks(np.linspace(0, 8, 3))
ax.plot([0, 0], [0, 12], 'r--', alpha=.3)
plf.subplottext(['A', 'B', 'C', 'D', 'E'][i], ax, x=-.08)
plf.spineless(ax)
print(i, stats.wilcoxon(group)[1])
plt.subplots_adjust(hspace=.3)
ax.set_xlabel(r'CSI$_{photopic}$ - CSI$_{mesopic}$')
texplot.savefig('csihistogram_mp')
plt.show()
manwhitu = stats.mannwhitneyu(csichange_on, csichange_off)
print('Mann-Whitney U test for ON and OFF populations '
f'being different: {manwhitu[1]:5.4f}')
# Function to quickly adjust the spine color and width
def setspines(ax):
locs = ['top', 'left', 'bottom', 'right']
for loc in locs:
ax.spines[loc].set_color('maroon')
ax.spines[loc].set_linewidth(1)
# Define functions to generate the representative stimuli
def checkerstim():
return np.random.randint(0, 2, bwndim*bwndim).reshape(bwndim,bwndim)
def barstim():
return (np.repeat(np.random.randint(0, 2, bwndim),
bwndim)).reshape(bwndim, bwndim)
# We first draw checker representation, then bar representation
# Since the bar should be further down, we offset it
for stim, baroffset in zip([checkerstim, barstim], [0, .45]):
np.random.seed(0)
for i in range(nr_frames):
d = i*offset # How much to offset each individual frame
# This is the trick to place the frames where you want
ax = fig.add_axes([.13-d, .7-d-baroffset, .06, .06])
ax.imshow(stim(), cmap='Greys')
plt.xticks([])
plt.yticks([])
setspines(ax)
plt.subplots_adjust(wspace=.3, hspace=.35)
texplot.savefig('comparesurroundcalc')
plt.show()
plt.close()
ax.set_xlim(ax_lims)
ax.set_ylim(ax_lims)
tick_locs = [0, 0.2, 0.4]
ax.set_xticks(tick_locs)
ax.set_yticks(tick_locs)
ax.scatter(csi[0, :], csi[1, :], s=6, c='grey')
ax.scatter(csi[0, group], csi[1, group], s=6, c=color, linewidths=.5)
ax.plot(x, x, 'k--', alpha=.5, linewidth=.5)
plf.subplottext(['B', 'C', 'D', 'E', 'F'][i], ax, x=-.25)
ax.set_aspect('equal')
plf.spineless(ax, 'tr')
plt.subplots_adjust(hspace=.4, wspace=.4)
texplot.savefig('csichange')
plt.show()
# Perform non-parametric paired test to test significance
stat_test = stats.wilcoxon(csi[0, :], csi[1, :])
print(f'Wilcoxon signed-rank test p-val: {stat_test[1]:7.2e}')
for label, group in zip(colorlabels, groups):
stat_test_grp = stats.wilcoxon(csi[0, group], csi[1, group])
print(f'Wilcoxon p-val for {label:25s}: {stat_test_grp[1]:7.2e}')
#np.savez('/home/ycan/Documents/thesis/analysis_auxillary_files/thesis_csiplotting.npz',
# cells=cells,
# bias=bias,
# include=include,
# colors=colors,
ax1 = axes[2*j]
plf.subplottext(['A', 'C'][j], ax1, x=-.4)
plf.subplottext(['Mesopic', 'Photopic'][j],
ax1, x=-.5, y=.5, rotation=90, va='center')
plf.stashow(sta, ax1, extent=[0, t[-1], -vscale, vscale],
cmap=texplot.cmap)
ax1.set_xlabel('Time [ms]')
ax1.set_ylabel(r'Distance [$\upmu$m]')
fitv = np.mean(sta[:, max_i[1]-cut_time:max_i[1]+cut_time+1],
axis=1)
s = np.arange(fitv.shape[0])
ax2 = axes[2*j+1]
plf.subplottext(['B', 'D'][j], ax2, x=-.1)
plf.subplottext(f'Center-Surround Index: {csi:4.2f}',
ax2, x=.95, y=.15, fontsize=8, fontweight='normal')
plf.spineless(ax2)
ax2.set_yticks([])
ax2.set_xticks([])
ax2.plot(onoroff*fitv, -s, label='Data')
# Displace the center of both distributions according to the difference
popt = popt - np.array([0, 1, 0, 0, 1, 0])*fsize_diff*2
ax2.plot(onedgauss(s, *popt[:3]), -s, '--', label='Center')
ax2.plot(-onedgauss(s, *popt[3:]), -s, '--', label='Surround')
plt.subplots_adjust(hspace = .3, wspace=.2)
texplot.savefig(label)
plt.show()
axes = fig.subplots(len(colorcategories) + 1, 1, sharex=True)
bins = np.arange(-200, 25 + 25,
25) # Arange does not include endpoint so 25+25
changes = []
for i, color in enumerate(colorcategories):
group = [index for index, c in enumerate(colors) if c == color]
ax = axes[i]
change = maxts_f[1, group] - maxts_f[0, group]
changes.append(change)
ax.hist(change, color=color, bins=bins)
ax.set_ylim([0, 10])
plf.subplottext(['A', 'B', 'C', 'D', 'E'][i], ax, x=-.08)
plf.spineless(ax)
stat_test = stats.wilcoxon(change)
print(f'Wilcoxon p-val for {color:25s}: {stat_test[1]:7.2e}')
ax = axes[-1]
ax.hist(maxts_f[1, :] - maxts_f[0, :], color='k', bins=bins)
plf.subplottext('F', ax, x=-.08)
plf.spineless(ax)
ax.set_xlabel(r'TTP\textsubscript{photopic} - TTP\textsubscript{mesopic} [ms]')
plt.subplots_adjust(hspace=.4)
texplot.savefig('temporalpeakchangehistogram')
plt.show()
mannwu = stats.mannwhitneyu(changes[0], changes[1])
print(f'Mann-Whitney-U test p-val for ON vs OFF categories: {mannwu[1]:7.2e}')
stats = stats.wilcoxon(maxts_f[1, :] - maxts_f[0, :])
print(f'Wilcoxon p-val for the whole population: {stats[1]:7.2e}')