def plot_regression(ax, x, y, xline, labels, number, **kwargs):
cov = covariance_package()
yline = np.zeros(len(xline))
ax.scatter(x,
y,
s=0.5,
color=kwargs['color'],
alpha=kwargs['alpha'][number])
if kwargs['fit_model'][number] == 'Local':
for i in range(len(xline)):
w = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i],
sig=GaussianWidth)
intercept, slope, sig = cov.linear_regression(x, y, weight=w)
yline[i] = slope * xline[i] + intercept
elif kwargs['fit_model'][number] == 'Global':
intercept, slope, sig = cov.linear_regression(x, y)
yline = slope * xline + intercept
ax.plot(xline,
yline,
color=kwargs['color'],
linestyle='-',
linewidth=1.3,
label=labels)
def run_number_count_model(self, x, xline, y, yline, xp=14, yp=13, deg=2):
cov = covariance_package()
nObs, dx = self.number_count_vector(y, yline)
print("Number of Halos : %i"%sum(nObs))
yObs = (yline[1:] + yline[:-1]) / 2.0
slope = np.zeros(len(yObs))
norm = np.zeros(len(yObs))
sig = np.zeros(len(yObs))
for i in range(len(yObs)):
w = cov.calculate_weigth(y, weigth_type='gussian', mu=yObs[i], sig=GaussianWidth)
norm[i], slope[i], sig[i] = cov.linear_regression(y, (x - xp), weight=w)
# Beild the model
model = self.likelihood_model1(nObs, yObs, slope, norm, sig, dx, xp=14, deg=deg)
# RUN MCMC
# do the MCMC sampling 10000 times and save all steps after 1000 initial steps
M = pymc.MCMC(model)
M.sample(iter=40000, burn=10000, verbose=0)
# save posterior distribution
self.print_results(M, x, xline, xp, deg=3, model_id=5)
self.print_inf_MF(M, x, xline, xp, deg=deg, model_id=5)
self.print_fit(M, x, xline, xp, deg=deg, model_id=5)
def scaling_parameters_at(self, x, y, xe=1.0, GaussianWidth=1.0):
cov = covariance_package()
w = cov.calculate_weigth(x, weigth_type='gussian', mu=xe, sig=GaussianWidth)
norm, slope, sig = cov.linear_regression(x, y, weight=w)
return slope, sig, norm
def plot_correlation_global(ax, x, y, z, xline, labels, number, **kwargs):
cov = covariance_package()
if labels == r'Z = 1 (BAHAMAS)':
return 0
if kwargs['fit_model'][number] == 'Global':
nmax = kwargs['nbins']
for i in range(nmax):
xmin = min(x) + float(i) * (max(x) - min(x)) / float(nmax)
xmax = min(x) + float(i + 1) * (max(x) - min(x)) / float(nmax)
mask = (x < xmax) * (x > xmin)
xp = x[mask]
yp = y[mask]
zp = z[mask]
lenX = len(xp)
cline = []
for iBoot in range(1000):
resample_i = np.floor(np.random.rand(lenX) * lenX).astype(int)
intercept, slope, _ = cov.linear_regression(
xp[resample_i], yp[resample_i])
dy = list(yp[resample_i] - slope * xp[resample_i] - intercept)
intercept, slope, _ = cov.linear_regression(
xp[resample_i], zp[resample_i])
dz = list(zp[resample_i] - slope * xp[resample_i] - intercept)
cline += [np.corrcoef(dy, dz)[1, 0]]
print labels, np.mean(xp), np.mean(cline), np.mean(xp), np.std(
cline)
# ax.plot(np.mean(xp), np.mean(cline), color=kwargs['color'], marker=kwargs['marker'],
# linestyle=None, linewidth=0, label=None)
ax.errorbar(np.mean(xp),
np.mean(cline),
yerr=np.std(cline),
color=kwargs['color'],
marker=kwargs['marker'],
label=None)
ax.plot(np.mean(xp),
np.mean(cline),
color=kwargs['color'],
marker=kwargs['marker'],
linestyle=None,
linewidth=0,
label=labels)
def scaling_parameters(self, x, y, GaussianWidth=1.0):
cov = covariance_package()
xline = np.linspace(np.min(x), np.max(x), 21)
xline = (xline[1:] + xline[:-1]) / 2.0
slope = np.zeros(len(xline))
norm = np.zeros(len(xline))
sig = np.zeros(len(xline))
for i in range(len(xline)):
w = cov.calculate_weigth(x, weigth_type='gussian', mu=xline[i], sig=GaussianWidth)
norm[i], slope[i], sig[i] = cov.linear_regression(x, y, weight=w)
return slope, sig, norm
def wu_calc(x, y, z, labels, number, **kwargs):
cov = covariance_package()
sig_y = []
sig_z = []
corr = []
if kwargs['fit_model'][number] == 'All':
print "Lable : ", labels
intercept, slope, _ = cov.linear_regression(x, y)
dy = list(y - slope * x - intercept)
print "Slope (Mgas , Mstar) : ", slope,
intercept, slope, _ = cov.linear_regression(x, z)
dz = list(z - slope * x - intercept)
print slope
corr = np.corrcoef(dy, dz)[1, 0]
print "Fractional scatter (Mgas , Mstar) : ", np.std(dy) * np.log(
10), np.std(dz) * np.log(10)
print "correlation ", corr
def plot_mass_evolution(ax, x, y, xline, labels, number, **kwargs):
cov = covariance_package()
if kwargs['fit_model'][number] == 'Local':
lenX = len(x)
slope = np.zeros([nBootstrap, len(xline)])
norm = np.zeros([nBootstrap, len(xline)])
sig = np.zeros([nBootstrap, len(xline)])
for iBoot in range(nBootstrap):
resample_i = np.floor(np.random.rand(lenX) * lenX).astype(int)
for i in range(len(xline)):
if kwargs['filter_type'] == 'gaussian':
w = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i],
sig=GaussianWidth)
elif kwargs['filter_type'] == 'uniform':
w = cov.calculate_weigth(x,
weigth_type='uniform',
xmin=xline[i] - .1,
xmax=xline[i] + 0.1)
norm[iBoot, i], slope[iBoot, i], sig[iBoot, i] =\
cov.linear_regression(x[resample_i], y[resample_i], weight=w[resample_i])
ax[0].plot(xline,
np.mean(slope, axis=0),
color=kwargs['color'],
linestyle='-',
linewidth=1.3,
label=labels)
ax[0].fill_between(xline,
np.percentile(slope, 16, axis=0),
np.percentile(slope, 84, axis=0),
facecolor=kwargs['color'],
alpha=0.4,
label=None)
ax[1].plot(xline,
np.log(10.0) * np.mean(sig, axis=0),
color=kwargs['color'],
linestyle='-',
linewidth=1.3,
label=labels)
ax[1].fill_between(xline,
np.log(10.0) * np.percentile(sig, 16, axis=0),
np.log(10.0) * np.percentile(sig, 84, axis=0),
facecolor=kwargs['color'],
alpha=0.4,
label=None)
ax[0].grid(True)
ax[1].grid(True)
elif kwargs['fit_model'][number] == 'Global':
lenX = len(x)
intercept = np.zeros(nBootstrap)
slope = np.zeros(nBootstrap)
sig = np.zeros(nBootstrap)
for iBoot in range(nBootstrap):
resample_i = np.floor(np.random.rand(lenX) * lenX).astype(int)
intercept[iBoot], slope[iBoot], sig[iBoot] = cov.linear_regression(
x[resample_i], y[resample_i])
ax[0].errorbar(np.mean(x),
np.mean(slope),
yerr=np.std(slope),
color=kwargs['color'],
marker=kwargs['marker'],
label=labels)
ax[1].errorbar(np.mean(x),
np.log(10.0) * np.mean(sig),
yerr=np.log(10.0) * np.std(sig),
color=kwargs['color'],
marker=kwargs['marker'],
label=labels)
def plot_cumulative_correlation(ax, x, y, z, xline, labels, number, **kwargs):
cov = covariance_package()
cline = np.zeros(len(xline) - 1)
dy = []
dz = []
if kwargs['fit_model'][number] == 'Local':
for i in range(len(xline) - 2, -1, -1):
mask = (x > xline[i]) * (x < xline[i + 1])
if kwargs['filter_type'] == 'gaussian':
w1 = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i],
sig=GaussianWidth)
w2 = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i + 1],
sig=GaussianWidth)
elif kwargs['filter_type'] == 'uniform':
w1 = cov.calculate_weigth(x,
weigth_type='uniform',
xmin=xline[i] - .1,
xmax=xline[i] + 0.1)
w2 = cov.calculate_weigth(x,
weigth_type='uniform',
xmin=xline[i + 1] - .1,
xmax=xline[i + 1] + 0.1)
intercept, slope, _ = cov.linear_regression(x, y, weight=w1)
yline1 = slope * xline[i] + intercept
intercept, slope, _ = cov.linear_regression(x, y, weight=w2)
yline2 = slope * xline[i + 1] + intercept
slope = (yline2 - yline1) / (xline[i + 1] - xline[i])
intercept = -slope * xline[i] + yline1
dy += list(y[mask] - slope * x[mask] - intercept)
intercept, slope, _ = cov.linear_regression(x, z, weight=w1)
yline1 = slope * xline[i] + intercept
intercept, slope, _ = cov.linear_regression(x, z, weight=w2)
yline2 = slope * xline[i + 1] + intercept
slope = (yline2 - yline1) / (xline[i + 1] - xline[i])
intercept = -slope * xline[i] + yline1
dz += list(z[mask] - slope * x[mask] - intercept)
cline[i] = np.corrcoef(dy, dz)[1, 0]
ax.plot((xline[1:] + xline[:-1]) / 2.,
cline,
color=kwargs['color'],
marker=kwargs['marker'],
linestyle=None,
linewidth=0,
label=labels)
if kwargs['fit_model'][number] == 'Global':
intercept, slope, _ = cov.linear_regression(x, y)
dy = list(y - slope * x - intercept)
intercept, slope, _ = cov.linear_regression(x, z)
dz = list(z - slope * x - intercept)
cline = np.corrcoef(dy, dz)[1, 0]
ax.plot(np.mean(x) - 0.3,
cline,
color=kwargs['color'],
marker=kwargs['marker'],
linestyle=None,
linewidth=0,
label=labels)
def plot_qq(ax, x, y, xline, labels, number, **kwargs):
cov = covariance_package()
dy = []
if kwargs['fit_model'][number] == 'Local':
for i in range(len(xline) - 1):
if kwargs['filter_type'] == 'gaussian':
w1 = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i],
sig=GaussianWidth)
w2 = cov.calculate_weigth(x,
weigth_type='gussian',
mu=xline[i + 1],
sig=GaussianWidth)
elif kwargs['filter_type'] == 'uniform':
w1 = cov.calculate_weigth(x,
weigth_type='uniform',
xmin=xline[i] - .1,
xmax=xline[i] + 0.1)
w2 = cov.calculate_weigth(x,
weigth_type='uniform',
xmin=xline[i + 1] - .1,
xmax=xline[i + 1] + 0.1)
intercept, slope, _ = cov.linear_regression(x, y, weight=w1)
yline1 = slope * xline[i] + intercept
intercept, slope, _ = cov.linear_regression(x, y, weight=w2)
yline2 = slope * xline[i + 1] + intercept
slope = (yline2 - yline1) / (xline[i + 1] - xline[i])
intercept = -slope * xline[i] + yline1
mask = (x > xline[i]) * (x < xline[i + 1])
std = np.std(y[mask] - slope * x[mask] - intercept)
dy += list((y[mask] - slope * x[mask] - intercept) / std)
elif kwargs['fit_model'][number] == 'Global':
intercept, slope, sig = cov.linear_regression(x, y)
std = np.std(y - slope * x - intercept)
dy = list((y - slope * x - intercept) / std)
res, bin_edges = np.histogram(dy,
bins=kwargs['bins'] / 4,
range=kwargs['xrange'],
normed=True)
Mean, Variance, Skewness, Kurtosis = stats.describe(dy)[2:]
dy.sort()
norm = npr.normal(Mean, 1, len(dy))
norm.sort()
ax.plot(norm, dy, ".", color=kwargs['color'])
ax.plot((-5, 5), (-5, 5), 'k--', linewidth=2)
def __init__(self):
self.cov = covariance_package()
self.cosmo = cosmo