# adjust eff to match to obs
parameters['eff'] = parameters['eff'] * dfss.lld.mean()/dfss.lldmod.mean()
variable_parameters_mu_prior = np.array(
[parameters[k] for k in variable_parameter_names])
fit_ret = emcee_deconvolve_tm.fit_parameters_to_obs(t, observed_counts=dfss.lld.values,
radon_conc=onemBq * 10 * np.ones(len(dfss)), # or, could be []
internal_airt_history=dfss.airt.values,
parameters=parameters,
variable_parameter_names = variable_parameter_names,
variable_parameters_mu_prior = variable_parameters_mu_prior,
variable_parameters_sigma_prior = variable_parameters_sigma_prior,
iterations=100,
thin=1,
keep_burn_in_samples=False,
nthreads=1)
(sampler, A, mean_est, low, high, parameters, map_radon_timeseries,
rl_radon_timeseries, rltv_radon_timeseries) = fit_ret
popt = A.mean(axis=0)
pmap = A[0,:]
print("(from bayes_fit...) pmap:", pmap)
params_chain = A[:, parameters['nstate']:parameters['nhyper']+parameters['nstate']]
radon_conc_chain = A[:, parameters['nhyper']+parameters['nstate']:]
expected_background*tres
f, ax = plt.subplots()
dfss[['lld','lldmod']].plot(ax=ax)
ax.set_title(dfss.index.to_pydatetime()[0].date().strftime('%b %Y'))
plt.show()
# fit to obs
if True:
fit_ret = emcee_deconvolve_tm.fit_parameters_to_obs(t, observed_counts=dfss.lld.values,
radon_conc=[], #dfss.cal_radon_conc.values/lamrn*0.0,
internal_airt_history=dfss.airt.values,
parameters=parameters,
variable_parameter_names = variable_parameter_names,
variable_parameters_mu_prior = variable_parameters_mu_prior,
variable_parameters_sigma_prior = variable_parameters_sigma_prior,
iterations=2000,
thin=10,
keep_burn_in_samples=False,
nthreads=1)
(sampler, A, mean_est, low, high, parameters, map_radon_timeseries,
rl_radon_timeseries, rltv_radon_timeseries) = fit_ret
popt = A.mean(axis=0)
pmap = A[0,:]
print("(from bayes_fit...) pmap:", pmap)
params_chain = A[:, parameters['nstate']:parameters['nhyper']+parameters['nstate']]
radon_conc_chain = A[:, parameters['nhyper']+parameters['nstate']:]
