def plot(a, parameters):
import matplotlib.pyplot as plt
p = pymultinest.PlotMarginal(a)
n_params = len(parameters)
values = a.get_equal_weighted_posterior()
s = a.get_stats()
assert n_params == len(s['marginals'])
prefix = a.outputfiles_basename
plt.figure(figsize=(5 * n_params, 5 * n_params))
for i in range(n_params):
plt.subplot(n_params, n_params, n_params * i + i + 1)
plot_marg(parameters, values, i, s, p)
for j in range(i):
plt.subplot(n_params, n_params, n_params * j + i + 1)
p.plot_conditional(i, j, bins=20, cmap=plt.cm.gray_r)
plt.xlabel(parameters[i])
plt.ylabel(parameters[j])
#plt.savefig('cond_%s_%s.pdf' % (params[i], params[j]), bbox_tight=True)
#plt.close()
plt.savefig(prefix + 'cond.pdf')
plt.close()
plt.figure(figsize=(4 * 4, 2 * math.ceil(n_params / 4.)))
for i in range(n_params):
plt.subplot(math.ceil(n_params / 4.), 4, i + 1)
plot_marg(parameters, values, i, s, p)
plt.savefig(prefix + 'marg.pdf', bbox_inches='tight')
plt.close()
sys.stderr.write("""SYNOPSIS: %s <output-root>
output-root: Where the output of a MultiNest run has been written to.
Example: chains/1-
%s""" % (sys.argv[0], __doc__))
sys.exit(1)
prefix = sys.argv[1]
parameters = json.load(file(prefix + 'params.json'))
n_params = len(parameters)
a = pymultinest.Analyzer(n_params=n_params, outputfiles_basename=prefix)
s = a.get_stats()
p = pymultinest.PlotMarginal(a)
values = a.get_equal_weighted_posterior()
assert n_params == len(s['marginals'])
modes = s['modes']
dim2 = os.environ.get('D', '1' if n_params > 20 else '2') == '2'
nbins = 100 if n_params < 3 else 20
if dim2:
plt.figure(figsize=(5 * n_params, 5 * n_params))
for i in range(n_params):
plt.subplot(n_params, n_params, i + 1)
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
def marginal_plots(analyzer, d=None):
"""
Create marginal plots
* analyzer: A instance of pymultinest.Analyzer
* d: if more than 20 dimensions, by default only 1d marginal distributions
are plotted. set d=2 if you want to force a 2d matrix plot
"""
prefix = analyzer.outputfiles_basename
n_params = analyzer.n_params
parameters = json.load(open(prefix + 'params.json'))
a = analyzer
s = a.get_stats()
p = pymultinest.PlotMarginal(a)
values = a.get_equal_weighted_posterior()
assert n_params == len(s['marginals'])
modes = s['modes']
dim2 = ((1 if n_params > 20 else 2) if d is None else d) == 2
if dim2:
plt.figure(figsize=(5 * n_params, 5 * n_params))
for i in range(n_params):
plt.subplot(n_params, n_params, i + 1)
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=20,
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
#print center, low1, high1
newax.errorbar(x=center,
y=y,
xerr=numpy.transpose(
[[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
#newax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(m['5sigma'])
oldax.set_xlim(m['5sigma'])
#plt.close()
for j in range(i):
plt.subplot(n_params, n_params, n_params * (j + 1) + i + 1)
p.plot_conditional(i, j, bins=20, cmap=plt.cm.gray_r)
for m in modes:
plt.errorbar(x=m['mean'][i],
y=m['mean'][j],
xerr=m['sigma'][i],
yerr=m['sigma'][j])
plt.xlabel(parameters[i])
plt.ylabel(parameters[j])
plt.xlim(s['marginals'][i]['5sigma'])
plt.ylim(s['marginals'][j]['5sigma'])
#plt.savefig('cond_%s_%s.pdf' % (params[i], params[j]), bbox_tight=True)
#plt.close()
plt.savefig(prefix + 'marg.pdf')
plt.savefig(prefix + 'marg.png')
plt.close()
else:
from matplotlib.backends.backend_pdf import PdfPages
print('1dimensional only. Set the D environment variable D=2 to force')
print('2d marginal plots.')
pp = PdfPages(prefix + 'marg1d.pdf')
for i in range(n_params):
plt.figure(figsize=(5, 5))
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=20,
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
print(center, low1, high1)
newax.errorbar(x=center,
y=y,
xerr=numpy.transpose(
[[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(m['5sigma'])
oldax.set_xlim(m['5sigma'])
plt.savefig(pp, format='pdf', bbox_inches='tight')
plt.close()
pp.close()
def mkplots(self):
# run to make plots of the resulting posteriors. Modified from marginal_plots.py
# from pymultinest. Produces basename+marg.pdf and basename+marge.png files
prefix = self.basename
parameters = json.load(file(prefix + 'params.json'))
n_params = len(parameters)
a = pymultinest.Analyzer(n_params=n_params,
outputfiles_basename=prefix)
s = a.get_stats()
p = pymultinest.PlotMarginal(a)
try:
values = a.get_equal_weighted_posterior()
except IOError as e:
print 'Unable to open: %s' % e
return
assert n_params == len(s['marginals'])
modes = s['modes']
dim2 = os.environ.get('D', '1' if n_params > 20 else '2') == '2'
nbins = 100 if n_params < 3 else 20
if dim2:
plt.figure(figsize=(5.1 * n_params, 5 * n_params))
for i in range(n_params):
plt.subplot(n_params, n_params, i + 1)
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=nbins,
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
print center, low1, high1
newax.errorbar(x=center,
y=y,
xerr=np.transpose(
[[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
#newax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(m['5sigma'])
oldax.set_xlim(m['5sigma'])
#plt.close()
for j in range(i):
plt.subplot(n_params, n_params, n_params * (j + 1) + i + 1)
p.plot_conditional(i, j, bins=20, cmap=plt.cm.gray_r)
for m in modes:
plt.errorbar(x=m['mean'][i],
y=m['mean'][j],
xerr=m['sigma'][i],
yerr=m['sigma'][j])
ax = plt.gca()
if j == i - 1:
plt.xlabel(parameters[i])
plt.ylabel(parameters[j])
[l.set_rotation(45) for l in ax.get_xticklabels()]
else:
ax.set_xticklabels([])
ax.set_yticklabels([])
plt.xlim([
m['mean'][i] - 5 * m['sigma'][i],
m['mean'][i] + 5 * m['sigma'][i]
])
plt.ylim([
m['mean'][j] - 5 * m['sigma'][j],
m['mean'][j] + 5 * m['sigma'][j]
])
#plt.savefig('cond_%s_%s.pdf' % (params[i], params[j]), bbox_tight=True)
#plt.close()
plt.tight_layout()
plt.savefig(prefix + 'marg.pdf')
plt.savefig(prefix + 'marg.png')
plt.close()
else:
from matplotlib.backends.backend_pdf import PdfPages
print '1dimensional only. Set the D environment variable D=2 to force'
print '2d marginal plots.'
pp = PdfPages(prefix + 'marg1d.pdf')
for i in range(n_params):
plt.figure(figsize=(5, 5))
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=20,
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
print center, low1, high1
newax.errorbar(x=center,
y=y,
xerr=np.transpose(
[[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(m['5sigma'])
oldax.set_xlim(m['5sigma'])
plt.savefig(pp, format='pdf', bbox_inches='tight')
plt.close()
pp.close()
def plot2D(a, parameters, outputDirectory, outputPrefix, nbins=20):
'''
Helper function for plotting 2D distributions from PyMultiNest.
The arguments are:
a - A PyMultiNest Analyzer object.
parameters - A list containing the names of all parameters.
outputDirectory - The directory in which to place output files.
outputPrefix - The prefix for all output file names.
nbins - The number of bins to use when plotting. Default is 20.
There is no return value, as all output is written to files.
'''
prefix = outputDirectory + '/' + outputPrefix
n_params = len(parameters)
p = pymultinest.PlotMarginal(a)
values = a.get_equal_weighted_posterior()
s = a.get_stats()
modes = s['modes']
plt.figure(figsize=(2.5 * n_params, 2.5 * n_params))
for i in range(n_params):
plt.subplot(n_params, n_params, i + 1)
plt.xlabel(parameters[i])
m = s['marginals'][i]
plt.xlim(m['5sigma'])
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=nbins,
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
#print(center, low1, high1)
newax.errorbar(x=center,
y=y,
xerr=np.transpose([[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
#newax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(m['5sigma'])
oldax.set_xlim(m['5sigma'])
#plt.close()
for j in range(i):
plt.subplot(n_params, n_params, n_params * (j + 1) + i + 1)
p.plot_conditional(i, j, bins=20, cmap=plt.cm.gray_r)
for m in modes:
plt.errorbar(x=m['mean'][i],
y=m['mean'][j],
xerr=m['sigma'][i],
yerr=m['sigma'][j])
plt.xlabel(parameters[i])
plt.ylabel(parameters[j])
#plt.savefig('cond_%s_%s.pdf' % (params[i], params[j]), bbox_tight=True)
#plt.close()
plt.tight_layout(rect=[0.07, 0.03, 0.97, 0.97], pad=0.2)
plt.savefig(prefix + 'marg.pdf')
plt.savefig(prefix + 'marg.png')
plt.close()
def plot1D(a, parameters, outputDirectory, outputPrefix):
'''
Helper function for plotting 1D distributions from PyMultiNest.
The arguments are:
a - A PyMultiNest Analyzer object.
parameters - A list containing the names of all parameters.
outputDirectory - The directory in which to place output files.
outputPrefix - The prefix for all output file names.
There is no return value, as all output is written to files.
'''
prefix = outputDirectory + '/' + outputPrefix
n_params = len(parameters)
p = pymultinest.PlotMarginal(a)
values = a.get_equal_weighted_posterior()
s = a.get_stats()
modes = s['modes']
pp = PdfPages(prefix + 'marg1d.pdf')
for i in range(n_params):
plt.figure(figsize=(3, 3))
plt.xlabel(parameters[i])
plt.locator_params(nbins=5)
m = s['marginals'][i]
iqr = m['q99%'] - m['q01%']
xlim = m['q01%'] - 0.3 * iqr, m['q99%'] + 0.3 * iqr
#xlim = m['5sigma']
plt.xlim(xlim)
oldax = plt.gca()
x, w, patches = oldax.hist(values[:, i],
bins=np.linspace(xlim[0], xlim[1], 20),
edgecolor='grey',
color='grey',
histtype='stepfilled',
alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(),
sharex=oldax,
frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05 * (ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
#print center, low1, high1
newax.errorbar(x=center,
y=y,
xerr=np.transpose([[center - low1, high1 - center]]),
color='blue',
linewidth=2,
marker='s')
oldax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(xlim)
oldax.set_xlim(xlim)
plt.savefig(pp, format='pdf', bbox_inches='tight')
plt.close()
pp.close()
def marginals(prefix, params):
print('model "%s"' % prefix)
parameters = json.load(open(prefix + 'params.json'))
n_params = len(parameters)
# Get posterior data
a = pymultinest.Analyzer(n_params = n_params, outputfiles_basename = prefix)
s = a.get_stats()
json.dump(s, open(prefix + 'stats.json', 'w'), indent=4)
# Print posterior data
print(' marginal likelihood:')
print(' ln Z = %.1f +- %.1f' % (s['global evidence'], s['global evidence error']))
print(' parameters:')
# Loop across parameters, obtain median and std.dev
for p, m in zip(parameters, s['marginals']):
lo, hi = m['1sigma']
med = m['median']
sigma = (hi - lo) / 2
i = max(0, int(-numpy.floor(numpy.log10(sigma))) + 1)
fmt = '%%.%df' % i
fmts = '\t'.join([' %-15s' + fmt + " +- " + fmt])
print(fmts % (p, med, sigma))
# Initialize PlotMarginal object
print('creating marginal plot ...')
p = pymultinest.PlotMarginal(a)
# Get posterior and mode info
values = a.get_equal_weighted_posterior()
assert n_params == len(s['marginals'])
modes = s['modes']
# Set $D = 1 if more than 20 parameters, 2 otherwise
dim2 = os.environ.get('D', '1' if n_params > 20 else '2') == '2'
nbins = 100 if n_params < 3 else 20
# If $D==2...
if dim2:
# Initialize plot
fig = plt.figure(figsize=(5*n_params, 5*n_params))
# Loop across parameters, create 1D plot
counter = 0
for key in params.keys():
for i in range(int(params[key]['nparams'])):
par = params[key]['param'+str(i+1)]
sp = par.split(',')
ax1 = fig.add_subplot(n_params, n_params, n_params*i+(i+1))
ax1.set_xlabel(parameters[i].replace('_','\_'))
# Get parameter statistics, set bounds to +/- 5sigma
m = s['marginals'][i]
# Histogram posterior
x,w,patches = ax1.hist(values[:,i], bins=nbins, edgecolor='black', color='mediumslateblue',\
histtype='stepfilled', alpha=0.2,normed=True)
if sp[2][:5] == 'param' or sp[3][:5]=='param':
print('not printing prior...')
else:
add_prior(ax1,sp[1],float(sp[2]),float(sp[3]))
ax1.set_ylim(0, x.max())
# y-value at which to place error-bar (1/20th of vertical extent)
ylim = ax1.get_ylim()
y = ylim[0] + 0.05*(ylim[1] - ylim[0])
# Place error-bar
center = m['median']
low1, high1 = m['1sigma']
ax1.errorbar(x=center, y=y,
xerr=numpy.transpose([[center - low1, high1 - center]]),
color='blue', linewidth=2, marker='s')
# Axis formatting
ax1.set_ylabel("PDF")
#ax1.set_xlim(m['5sigma'])
# Loop across other parameters
for j in range(i+1,n_params):
# Set up 2D posterior plot
ax2 = fig.add_subplot(n_params, n_params, n_params*j+(i+1),sharex=ax1)
p.plot_conditional(i, j, cmap = 'Blues', bins='log')
# Plot error bars for each mode
for m in modes:
ax2.errorbar(x=m['mean'][i], y=m['mean'][j], xerr=m['sigma'][i],\
yerr=m['sigma'][j])
# Axes labels
ax2.set_xlabel(parameters[i].replace('_','\_'))
ax2.set_ylabel(parameters[j].replace('_','\_'))
# Save figure
plt.savefig(prefix + 'marg.pdf')
plt.savefig(prefix + 'marg.png')
plt.close()
# If $D==1...
else:
# Print stuff
from matplotlib.backends.backend_pdf import PdfPages
sys.stderr.write('1dimensional only. Set the D environment variable \n')
sys.stderr.write('to D=2 to force 2d marginal plots.\n')
pp = PdfPages(prefix + 'marg1d.pdf')
# Proceed as above, but only plot 1D posterior plots
for i in range(n_params):
plt.figure(figsize=(3, 3))
plt.xlabel(parameters[i].replace('_','\_'))
plt.locator_params(nbins=5)
m = s['marginals'][i]
iqr = m['q99%'] - m['q01%']
xlim = m['q01%'] - 0.3 * iqr, m['q99%'] + 0.3 * iqr
#xlim = m['5sigma']
plt.xlim(xlim)
oldax = plt.gca()
x,w,patches = oldax.hist(values[:,i],
bins=numpy.linspace(xlim[0], xlim[1], 20),
edgecolor='black', color='grey', histtype='stepfilled', alpha=0.2)
oldax.set_ylim(0, x.max())
newax = plt.gcf().add_axes(oldax.get_position(), sharex=oldax, frameon=False)
p.plot_marginal(i, ls='-', color='blue', linewidth=3)
newax.set_ylim(0, 1)
ylim = newax.get_ylim()
y = ylim[0] + 0.05*(ylim[1] - ylim[0])
center = m['median']
low1, high1 = m['1sigma']
#print center, low1, high1
newax.errorbar(x=center, y=y,
xerr=numpy.transpose([[center - low1, high1 - center]]),
color='blue', linewidth=2, marker='s')
oldax.set_yticks([])
newax.set_ylabel("Probability")
ylim = oldax.get_ylim()
newax.set_xlim(xlim)
oldax.set_xlim(xlim)
plt.savefig(pp, format='pdf', bbox_inches='tight')
plt.close()
pp.close()
