def plot_bestmodel(self, fitter, modelcode, folded=False, burnin_pct=0.1):
print('calling _mp_visuals.py/plot_bestmodel().')
### THIS FUNCTION PLOTS YOUR BEST FIT LIGHT CURVE MODEL OVER THE DATA.
if folded == True:
self.fold()
if modelcode == "LUNA":
folded = False ### should not be generating a folded light curve for a moon fit.
self.initialize_priors(modelcode=modelcode)
if fitter == 'emcee':
if modelcode == 'batman':
chainsdir = moonpydir + '/emcee_fits/batman/' + str(
self.target) + '/chains'
elif modelcode == 'LUNA':
chainsdir = moonpydir + '/emcee_fits/LUNA/' + str(
self.target) + '/chains'
samples = np.genfromtxt(chainsdir + '/' + str(self.target) +
'_mcmc_chain.txt')
sample_shape = samples.shape
samples = samples[int(burnin_pct * sample_shape[0]):, 1:]
best_fit_dict = {}
for npar, parlab in enumerate(self.param_labels):
best_fit_dict[parlab] = np.nanmedian(samples.T[npar])
print("best fit values: ")
for parkey in best_fit_dict.keys():
print(parkey, ' = ', best_fit_dict[parkey])
if modelcode == 'batman':
### use batman to generate a model!!!
if folded == True:
batman_times, batman_fluxes = run_batman(self.fold_times,
**best_fit_dict,
add_noise='n',
show_plots='n')
plt.scatter(np.hstack(self.fold_times),
np.hstack(self.fluxes_detrend),
facecolor='LightCoral',
edgecolor='k')
else:
batman_times, batman_fluxes = run_batman(self.times,
**best_fit_dict,
add_noise='n',
show_plots='n')
plt.scatter(np.hstack(self.times),
np.hstack(self.fluxes_detrend),
facecolor='LightCoral',
edgecolor='k')
batman_sort = np.argsort(batman_times)
batman_times, batman_fluxes = batman_times[
batman_sort], batman_fluxes[batman_sort]
plt.plot(batman_times, batman_fluxes, c='g', linewidth=2)
plt.show()
def pymn_loglike_batman(cube, ndim, nparams):
pymn_var_dict = {
} #### dictionary of variables, will take the cube as the argument.
pymn_fixed_dict = {
} ### dictionary of fixed variables, will stay constant for each run.
for pidx, parlab in enumerate(mn_variable_labels):
pymn_var_dict[parlab] = cube[pidx]
for pidx, parlab in enumerate(mn_fixed_labels):
pymn_fixed_dict[parlab] = mn_param_dict[parlab][
1] ### grabs the fixed value!
### now you should be able to run_LUNA(param_dict)
batman_times, batman_fluxes = run_batman(data_times,
**pymn_var_dict,
**pymn_fixed_dict,
model=mn_model,
add_noise='n',
show_plots='n')
loglikelihood = np.nansum(
-0.5 * ((batman_fluxes - data_fluxes) / data_errors)**2
) ### SHOULD MAKE THIS BETTER, to super-penalize running out of bounds!
return loglikelihood
def emcee_lnlike_batman(params, data_times, data_fluxes, data_errors):
emcee_param_dict = {} ### initialize it
emcee_var_param_dict = {}
emcee_fixed_param_dict = {}
"""
for pidx, parlab in enumerate(mn_param_labels):
emcee_param_dict[parlab] = params[pidx]
"""
for pidx, parlab in enumerate(mc_variable_labels):
emcee_var_param_dict[parlab] = params[pidx]
for pidx, parlab in enumerate(mc_fixed_labels):
emcee_fixed_param_dict[parlab] = mc_param_dict[parlab][
1] ### fixed value!
batman_times, batman_fluxes = run_batman(data_times,
**emcee_var_param_dict,
**emcee_fixed_param_dict,
model=mc_model,
add_noise='n',
show_plots='n')
loglikelihood = np.nansum(
-0.5 * ((batman_fluxes - data_fluxes) / data_errors)**2
) ### SHOULD MAKE THIS BETTER, to super-penalize running out of bounds!
#if (np.isfinite(loglikelihood) == False) or type(loglikelihood) != float:
# print("loglikelihood = ", loglikelihood)
# raise Exception('emcee_lnlike_batman loglikelihood value is invalid."')
return loglikelihood