def quantiles(fit):
tab = fit.load_mnest_results()
pars = tab.colnames
weights = tab['weights']
sumweights = np.sum(weights)
weights = weights / sumweights
sig1 = 0.682689
sig2 = 0.9545
sig3 = 0.9973
sig1_lo = (1. - sig1) / 2.
sig2_lo = (1. - sig2) / 2.
sig3_lo = (1. - sig3) / 2.
sig1_hi = 1. - sig1_lo
sig2_hi = 1. - sig2_lo
sig3_hi = 1. - sig3_lo
# Calculate the median and quantiles.
med_vals = {}
for n in pars:
# Calculate median, 1 sigma lo, and 1 sigma hi credible interval.
med_vals[n] = model_fitter.weighted_quantile(tab[n],
[0.5, sig1_lo, sig1_hi],
sample_weight=weights)
# Switch from values to errors.
med_vals[n][1] = med_vals[n][0] - med_vals[n][1]
med_vals[n][2] = med_vals[n][2] - med_vals[n][0]
return pars, med_vals
def summarize_results(tab):
if len(tab) < 1:
print('Did you run multinest_utils.separate_mode_files yet?')
# Which params to include in table
parameters = tab.colnames
parameters.remove('weights')
parameters.remove('logLike')
weights = tab['weights']
sumweights = np.sum(weights)
weights = weights / sumweights
sig1 = 0.682689
sig2 = 0.9545
sig3 = 0.9973
sig1_lo = (1.-sig1)/2.
sig2_lo = (1.-sig2)/2.
sig3_lo = (1.-sig3)/2.
sig1_hi = 1.-sig1_lo
sig2_hi = 1.-sig2_lo
sig3_hi = 1.-sig3_lo
print(sig1_lo, sig1_hi)
# Calculate the median, best-fit, and quantiles.
best_idx = np.argmax(tab['logLike'])
best = tab[best_idx]
best_errors = {}
med_best = {}
med_errors = {}
for n in parameters:
# Calculate median, 1 sigma lo, and 1 sigma hi credible interval.
tmp = model_fitter.weighted_quantile(tab[n], [0.5, sig1_lo, sig1_hi], sample_weight=weights)
# Switch from values to errors.
err_lo = tmp[0] - tmp[1]
err_hi = tmp[2] - tmp[0]
# Store into dictionaries.
med_best[n] = tmp[0]
med_errors[n] = np.array([err_lo, err_hi])
#best_errors[n] = np.array([best[n] - tmp[1], tmp[2] - best[n]])
best_errors[n] = np.array([tmp[1], tmp[2]])
print('####################')
print('Best-Fit Solution:')
print('####################')
fmt = ' {0:15s} best = {1:10.3f} 68\% low = {2:10.3f} 68% hi = {3:10.3f}'
for n in parameters:
print(fmt.format(n, best[n], best_errors[n][0], best_errors[n][1]))
return
def table_phot_astrom():
"""
I was in a rush and didn't have time to make a proper table, so this just
prints out the values, and you manually type it into LaTeX itself.
"""
#####
# Comment out the target you aren't working on
#####
target = 'MB10364'
target_name = 'mb10364'
# target='OB110462'
# target_name='ob110462'
mod_fit, data = lu_2019_lens.get_data_and_fitter(tdir[target_name])
mod_all = mod_fit.get_best_fit_modes_model(def_best='maxL')
mod = mod_all[0]
tab = mod_fit.load_mnest_results()
tab['piE'] = np.hypot(tab['piE_E'], tab['piE_N'])
tab['muRel'] = np.hypot(tab['muRel_E'], tab['muRel_N'])
tab_best = {}
tab_maxl = {}
med_errors = {}
sumweights = np.sum(tab['weights'])
weights = tab['weights'] / sumweights
sig1 = 0.682689
sig2 = 0.9545
sig3 = 0.9973
sig1_lo = (1. - sig1) / 2.
sig2_lo = (1. - sig2) / 2.
sig3_lo = (1. - sig3) / 2.
sig1_hi = 1. - sig1_lo
sig2_hi = 1. - sig2_lo
sig3_hi = 1. - sig3_lo
#####
# Comment out the target you aren't working on
#####
# For MB10364
params = ['tE', 'thetaE', 'piE', 'muRel', 'mL', 'mag_src2', 'piL']
# For OB110462
# params = ['tE', 'thetaE', 'piE', 'muRel', 'mL', 'mag_src3', 'piL']
maxl_idx = tab['logLike'].argmax()
for n in params:
tab_maxl[n] = tab[n][maxl_idx]
# Calculate median, 1 sigma lo, and 1 sigma hi credible interval.
tmp = model_fitter.weighted_quantile(tab[n], [0.5, sig1_lo, sig1_hi],
sample_weight=weights)
tab_best[n] = tmp[0]
# Switch from values to errors.
# err_lo = tmp[0] - tmp[1]
# err_hi = tmp[2] - tmp[0]
# med_errors[n] = np.array([err_lo, err_hi])
err_lo = tab_maxl[n] - tmp[1]
err_hi = tmp[2] - tab_maxl[n]
med_errors[n] = np.array([err_lo, err_hi])
print(tab_maxl)
print(med_errors)
return