def extract_posteriors():
logMs, logZs, dusts, t0s, logTaus = [], [], [], [], []
clusters = ['a370', 'a1063', 'a2744', 'm416', 'm717', 'm1149']
for cluster in clusters:
h5_files = glob.glob('{}/h5/*.h5'.format(cluster))
h5s = []
for file in h5_files:
file = file.replace(os.sep, '/') # compatibility for Windows
h5s.append(file)
for file in h5s:
result, obs, _ = reader.results_from(file, dangerous=True)
samples = sample_posterior(result['chain'],
weights=result['weights'],
nsample=len(result['chain']))
samples[:, 1] = np.log10(samples[:, 1])
samples[:, 5] = np.log10(samples[:, 5])
logMs.append(list(samples[:, 1]))
logZs.append(list(samples[:, 2]))
dusts.append(list(samples[:, 3]))
t0s.append(list(samples[:, 4]))
logTaus.append(list(samples[:, 5]))
logM = flatten_posteriors(logMs)
logZ = flatten_posteriors(logZs)
dust = flatten_posteriors(dusts)
t0 = flatten_posteriors(t0s)
logTau = flatten_posteriors(logTaus)
# np.savez('subsample_posteriors.npz', logM=logM, logZ=logZ, dust=dust,
# t0=t0, logTau=tau)
solMetal = u.def_unit(['solMetal', 'Z_sun', 'Zsun'],
prefixes=False,
format={
'latex': r'Z_{\odot}',
'unicode': 'Z\N{SUN}'
})
table = Table(
[logM / u.solMass, logZ / solMetal, dust, t0 * u.Gyr, logTau / u.Gyr],
names=('logM', 'logZ', 'dust', 't0', 'logTau'))
table.write('boneyard/subsample_posteriors.fits')
return
def determine_age_gradients(cluster, ID, bins, radius_array, version='') :
mwas = []
radii = []
for binNum, radius in zip(bins, radius_array) :
infile = '{}/h5/{}_ID_{}_bin_{}{}.h5'.format(
cluster, cluster, ID, binNum, version)
result, obs, _ = reader.results_from(infile, dangerous=True)
samples = sample_posterior(result['chain'], weights=result['weights'])
mwas.append(mwa_calc(samples[:, 5], samples[:, 4], power=1))
radii.append(np.full(len(samples), radius))
return np.concatenate(radii).ravel(), np.concatenate(mwas).ravel()
def save_mass_metallicity_images():
HFF = Table.read('output/tables/nbCGs.fits')
for cluster, ID in zip(HFF['cluster'], HFF['ID']):
# ensure the output directories for the images are available
os.makedirs('{}/logM_images'.format(cluster), exist_ok=True)
os.makedirs('{}/logZ_images'.format(cluster), exist_ok=True)
bins_image = core.open_image(cluster, ID, 'bins')
bins = np.sort(np.unique(
bins_image[~np.isnan(bins_image)])).astype(int)
mass = bins_image.copy()
metal = bins_image.copy()
if not (cluster == 'm717') & (ID == 3692): # fitting issue for bin_4
for binNum in bins: # loop over all the bins
result, obs, _ = reader.results_from(
'{}/h5/{}_ID_{}_bin_{}.h5'.format(cluster, cluster, ID,
binNum),
dangerous=True)
samples = sample_posterior(result['chain'],
weights=result['weights'])
mass[mass == binNum] = np.percentile(samples[:, 1], 50)
metal[metal == binNum] = np.percentile(samples[:, 2], 50)
hdu = fits.PrimaryHDU(mass)
hdu.writeto('{}/logM_images/{}_ID_{}_logM.fits'.format(
cluster, cluster, ID))
hdu = fits.PrimaryHDU(metal)
hdu.writeto('{}/logZ_images/{}_ID_{}_logZ.fits'.format(
cluster, cluster, ID))
return
#sys.exit(0)
results_type = 'dynesty'
result, obs, _ = reader.results_from(adap_dir + results_type + "_" + \
field + "_" + str(galaxy_seq) + ".h5", dangerous=False)
parnames = np.array(result['theta_labels'])
print('Parameters in this model:', parnames)
# ------------------
from prospect.plotting.utils import sample_posterior
wts = result.get("weights", None)
theta = sample_posterior(result["chain"], weights=wts, nsample=nsamp)
"""
fig1 = plt.figure()
ax1 = fig1.add_subplot(111)
ax1.set_xlabel(r'$\mathrm{Time\, [Gyr]}$', fontsize=13)
ax1.set_ylabel(r'$\mathrm{SFR\, [M_\odot/yr]}$', fontsize=13)
t = np.arange(0.0, 14.0, 0.001)
tau_arr = theta[:, -1]
mid_tau = np.mean(tau_arr)
min_tau = np.min(tau_arr)
max_tau = np.max(tau_arr)
def determine_lines(cluster, ID, bins, best=False, version='') :
metals_16, metals_50, metals_84 = [], [], []
dusts_16, dusts_50, dusts_84 = [], [], []
mwas_16, mwas_50, mwas_84 = [], [], []
metals_best, dusts_best, mwas_best = [], [], []
for binNum in bins : # loop over all the bins in the table
infile = '{}/h5/{}_ID_{}_bin_{}{}.h5'.format(
cluster, cluster, ID, binNum, version)
result, obs, _ = reader.results_from(infile, dangerous=True)
zred, mass, logzsol, dust2, tage, tau = result['bestfit']['parameter']
dusts_best.append(dust2)
metals_best.append(logzsol)
mwas_best.append(mwa_calc(tau, tage, power=1))
# now pull samples to compute 1sigma errors
samples = sample_posterior(result['chain'],
weights=result['weights'])
metal_16, metal_50, metal_84 = np.percentile(samples[:, 2],
[16, 50, 84])
metals_16.append(metal_16)
metals_50.append(metal_50)
metals_84.append(metal_84)
dust_16, dust_50, dust_84 = np.percentile(samples[:, 3],
[16, 50, 84])
dusts_16.append(dust_16)
dusts_50.append(dust_50)
dusts_84.append(dust_84)
mwa_16, mwa_50, mwa_84 = np.percentile(mwa_calc(samples[:, 5],
samples[:, 4],
power=1),
[16, 50, 84])
mwas_16.append(mwa_16)
mwas_50.append(mwa_50)
mwas_84.append(mwa_84)
metal_lo = np.abs(np.array(metals_50) - np.array(metals_16))
metal_hi = np.abs(np.array(metals_84) - np.array(metals_50))
dust_lo = np.abs(np.array(dusts_50) - np.array(dusts_16))
dust_hi = np.abs(np.array(dusts_84) - np.array(dusts_50))
mwa_lo = np.abs(np.array(mwas_50) - np.array(mwas_16))
mwa_hi = np.abs(np.array(mwas_84) - np.array(mwas_50))
metal_median = np.array(metals_50)
dust_median = np.array(dusts_50)
mwa_median = np.array(mwas_50)
metal_best = np.array(metals_best)
dust_best = np.array(dusts_best)
mwa_best = np.array(mwas_best)
if best :
return (metal_median, metal_lo, metal_hi,
dust_median, dust_lo, dust_hi,
mwa_median, mwa_lo, mwa_hi,
metal_best, dust_best, mwa_best)
else :
return (metal_median, metal_lo, metal_hi,
dust_median, dust_lo, dust_hi,
mwa_median, mwa_lo, mwa_hi)
mwas, mwa_16, mwa_84 = [], [], []
metals, metals_16, metals_84 = [], [], []
for binNum in bins: # loop over all the bins in the table
infile = '{}/h5/ID_{}_bin_{}.h5'.format(cluster, ID, binNum)
result, obs, _ = reader.results_from(infile, dangerous=True)
zred, mass, logzsol, dust2, tage, tau = result['bestfit'][
'parameter']
metals.append(logzsol)
mwa = mwa_calc(tau, tage, power=1)
mwas.append(mwa)
# print(mwa, mwa_calc_alt(tau, tage, power=1))
# print(cosmo.age(zred).value, tage, tau, mwa)
samples = sample_posterior(result['chain'],
weights=result['weights'])
sample_metals = samples[:, 2]
sample_mwas = mwa_calc(samples[:, 5], samples[:, 4], power=1)
mwa_16.append(np.percentile(sample_mwas, 16))
mwa_84.append(np.percentile(sample_mwas, 84))
metals_16.append(np.percentile(sample_metals, 16))
metals_84.append(np.percentile(sample_metals, 84))
mwa_lo = np.abs(mwas - np.array(mwa_16))
mwa_hi = np.abs(np.array(mwa_84) - mwas)
metals_lo = np.abs(metals - np.array(metals_16))
metals_hi = np.abs(np.array(metals_84) - metals)
def results_corner(cluster,
ID,
binNum,
full=False,
results_type='dynesty',
save=False,
version=''):
# create corner plots
result, obs, model, sps = get_results(cluster,
ID,
binNum,
results_type=results_type,
version=version)
if save:
os.makedirs(
'{}/images_corner_plots'.format(cluster), # ensure the output
exist_ok=True) # directory for the figures is available
outfile = '{}/images_corner_plots/{}_ID_{}_bin_{}{}.pdf'.format(
cluster, cluster, ID, binNum, version)
samples = sample_posterior(result['chain'], weights=result['weights'])
mwas = mwa_calc(samples[:, 5], samples[:, 4], power=1)
# logify the mass and tau parameters
samples[:, 1] = np.log10(samples[:, 1])
samples[:, 5] = np.log10(samples[:, 5])
samples = np.c_[samples, mwas]
z_lo, z_hi = np.percentile(samples[:, 0], [0.15, 99.85])
M_lo, M_hi = np.percentile(samples[:, 1], [0.15, 99.85])
Z_lo, Z_hi = np.percentile(samples[:, 2], [0.15, 99.85])
d_lo, d_hi = np.percentile(samples[:, 3], [0.15, 99.85])
t_lo, t_hi = np.percentile(samples[:, 4], [0.15, 99.85])
T_lo, T_hi = np.percentile(samples[:, 5], [0.15, 99.85])
MWA_lo, MWA_hi = np.percentile(samples[:, 6], [0.15, 99.85])
if full:
ranges = [(z_lo, z_hi), (M_lo, M_hi), (Z_lo, Z_hi), (d_lo, d_hi),
(t_lo, t_hi), (T_lo, T_hi), (MWA_lo, MWA_hi)]
labels = [
r'$z_{\rm red}$', r'$\log(M_{*})$', r'$\log(Z_{*})$',
r'$\hat{\tau}_{\lambda, 2}$', r'$T_{0}$', r'$\log(\tau)$', 'MWA'
]
plt.plot_corner(samples,
labels,
len(labels),
result,
ranges=ranges,
outfile=outfile,
save=save)
else:
sub_samples = np.delete(samples, [4, 5], 1)
sub_labels = [
r'$z_{\rm red}$', r'$\log(M_{*})$', r'$\log(Z_{*})$',
r'$\hat{\tau}_{\lambda, 2}$', 'MWA'
]
sub_ranges = [(z_lo, z_hi), (M_lo, M_hi), (Z_lo, Z_hi), (d_lo, d_hi),
(MWA_lo, MWA_hi)]
plt.plot_corner(sub_samples,
sub_labels,
len(sub_labels),
result,
ranges=sub_ranges,
outfile=outfile,
save=save)
return