def findHDCells(tuning_curves, ep, spikes, position, cut_off=0.49):
"""
Peak firing rate larger than 1
and Rayleigh test p<0.001 & z > 100
"""
cond1 = pd.DataFrame(tuning_curves.max() > 1.0)
angle = position.restrict(ep)
from pycircstat.tests import rayleigh
stat = pd.DataFrame(index=tuning_curves.columns, columns=['pval', 'z'])
for k in tuning_curves:
stat.loc[k] = rayleigh(tuning_curves[k].index.values,
tuning_curves[k].values)
#stat.loc[k]= rayleigh(position['ry'].restrict(ep).values , position['ry'].realign(spikes[k].restrict(ep)))
rMean = pd.DataFrame(index=tuning_curves.columns, columns=['hd_score'])
for k in tuning_curves:
"""computes the rayleigh vector length as hdScore.
"""
spk = spikes[k]
spk = spk.restrict(ep)
angle_spk = angle.realign(spk)
C = np.sum(np.cos(angle_spk.values))
S = np.sum(np.sin(angle_spk.values))
Rmean = np.sqrt(C**2 + S**2) / len(angle_spk)
rMean.loc[k] = Rmean
stat['hd_score'] = rMean
cond2 = pd.DataFrame(np.logical_and(stat['pval'] < 0.001, stat['z'] > 19))
cond3 = pd.DataFrame(rMean['hd_score'] >= cut_off)
'''To Do
Add cond 4 and set it to any value greater than 0.4'''
#cond4=pd.DataFrame(stat['hd_info'])
tokeep = (cond1[0] == True) & (cond2[0] == True) & (cond3['hd_score']
== True)
stat['hd_cells'] = tokeep
#tuning_curves have been normalised by occupancy, unlike the rvector
#Rayleigh z test to test the null hypothesis that there is no sample mean direction
'''I=pd.DataFrame(index=spikes.keys(),columns=['Ispk'])
for i in spikes.keys():
lamda_i=tuning_curves[i].values
bins=tuning_curves.index
lamda=len(spikes[i].restrict(ep))/ep.tot_length('s')
pos=position.restrict(ep)
bins=linspace(0,2*pi,60)
occu,a=np.histogram(pos, bins)
occu= occu/sum(occu)
bits_spk=sum(occu*(lamda_i/lamda)*np.log2(lamda_i/lamda))
I.loc[i,'Ispk']=bits_spk
stat['hd_info']=I'''
return stat
def findHDCells(tuning_curves):
"""
Peak firing rate larger than 1
and Rayleigh test p<0.001 & z > 100
"""
cond1 = tuning_curves.max() > 1.0
from pycircstat.tests import rayleigh
stat = pd.DataFrame(index=tuning_curves.columns, columns=['pval', 'z'])
for k in tuning_curves:
stat.loc[k] = rayleigh(tuning_curves[k].index.values,
tuning_curves[k].values)
cond2 = np.logical_and(stat['pval'] < 0.001, stat['z'] > 20)
tokeep = np.where(np.logical_and(cond1, cond2))[0]
return tokeep, stat
def findHDCells(tuning_curves, ep, spikes, position):
"""
Peak firing rate larger than 1
and Rayleigh test p<0.001 & z > 100
"""
cond1 = pd.DataFrame(tuning_curves.max() > 1.0)
angle = position.restrict(ep)
from pycircstat.tests import rayleigh
stat = pd.DataFrame(index=tuning_curves.columns, columns=['pval', 'z'])
for k in tuning_curves:
stat.loc[k] = rayleigh(tuning_curves[k].index.values,
tuning_curves[k].values)
#stat.loc[k]= rayleigh(position['ry'].restrict(ep).values , position['ry'].realign(spikes[k].restrict(ep)))
rMean = pd.DataFrame(index=tuning_curves.columns, columns=['hd_score'])
for k in tuning_curves:
"""computes the rayleigh vector length as hdScore.
"""
spk = spikes[k]
spk = spk.restrict(ep)
angle_spk = angle.realign(spk)
C = np.sum(np.cos(angle_spk.values))
S = np.sum(np.sin(angle_spk.values))
Rmean = np.sqrt(C**2 + S**2) / len(angle_spk)
rMean.loc[k] = Rmean
stat['hd_score'] = rMean
cond2 = pd.DataFrame(np.logical_and(stat['pval'] < 0.001, stat['z'] > 40))
cond3 = pd.DataFrame(rMean['hd_score'] >= 0.5)
tokeep = (cond1[0] == True) & (cond2[0] == True) & (cond3['hd_score']
== True)
stat['hd_cells'] = tokeep
def plot_correlogram(ax,
df,
tag,
direct='A->B',
overlap_key='overlap',
color='gray',
density=False,
regress=True,
x_dist=True,
y_dist=True,
ab_key='phaselag_AB',
ba_key='phaselag_BA',
alpha=0.2,
markersize=8,
linew=1):
"""
Parameters
----------
ax : matplotlib.axes._subplots.AxesSubplot
df : Dataframe
tag : str
direct : str
Either 'A->B', 'B->A' or 'combined'
overlap_key : str
Key of the overlap metric
color : str
Color of the scatter
regress : bool
x_dist : bool
y_dist : bool
ab_key : str
ba_key : str
alpha : float
Returns
-------
"""
overlap = df[overlap_key].to_numpy()
phaselag_AB = df[ab_key].to_numpy()
phaselag_BA = df[ba_key].to_numpy()
# Define x, y
if direct == 'A->B':
x = overlap
y = phaselag_AB
elif direct == 'B->A':
x = overlap
y = phaselag_BA
elif direct == 'combined':
x = np.concatenate([overlap, overlap])
y = np.concatenate([phaselag_AB, -phaselag_BA])
nan_mask = np.isnan(x) | np.isnan(y)
x_nonan, y_nonan = x[~nan_mask], y[~nan_mask]
# Scatter or density
if density:
xx, yy, zz = linear_circular_gauss_density(x_nonan,
y_nonan,
cir_kappa=3 * np.pi,
lin_std=0.05,
xbins=200,
ybins=800,
ybound=(-np.pi, 2 * np.pi))
ax.pcolormesh(xx, yy, zz)
else:
ax.scatter(x, y, c=color, alpha=alpha, s=markersize, marker='.')
ax.scatter(x,
y + (2 * np.pi),
c=color,
alpha=alpha,
s=markersize,
marker='.')
# Plot marginal mean
mean_y = circmean(y_nonan)
ax.plot([0, 0.2], [mean_y, mean_y], c='k', linewidth=linew)
# Regression
regress_d = None
if regress:
regress_d = rcc(x_nonan, y_nonan, abound=[-2, 2])
m, c, r, pval = regress_d['aopt'], regress_d['phi0'], regress_d[
'rho'], regress_d['p']
r_regress = np.linspace(0, 1, 20)
lag_regress = 2 * np.pi * r_regress * m
for tmpid, intercept in enumerate(
[2 * mul_idx * np.pi + c for mul_idx in [-2, -1, 0, 1, 2]]):
if tmpid == 0:
ax.plot(r_regress,
lag_regress + intercept,
c='k',
linewidth=linew,
label='rho=%0.2f (p%s)' % (r, p2str(pval)))
else:
ax.plot(r_regress,
lag_regress + intercept,
c='k',
linewidth=linew)
# X-axis
ax.plot([0, 1], [0, 0], c='k', alpha=1, linewidth=linew)
# (y-axis) Interpolation and Smoothened histrogram
if y_dist:
p, _ = rayleigh(y)
yax, yden = circular_density_1d(y_nonan, 10 * np.pi, 60,
(-np.pi, 3 * np.pi))
ax.plot(yden / yden.max() * 0.2, yax, c='k', linewidth=linew)
# (x-axis) Interpolation and Smoothened histrogram
if x_dist:
xax, xden = linear_density_1d(x_nonan,
std=0.05,
bins=100,
bound=(0, 1))
ax.plot(xax,
xden / xden.max() * (np.pi / 2) - np.pi,
c='k',
linewidth=linew)
# Set title
# ax.set_title("%s %s (n=%d)" % (tag, direct, num_pairs), fontsize=fontsize)
ax.set_ylim(-np.pi, 2 * np.pi)
ax.set_yticks([-np.pi, 0, np.pi, np.pi * 2])
_ = ax.set_yticklabels(['$-\pi$', '0', '$\pi$', '$2\pi$'])
return ax, x, y, regress_d
