def get_model_results_dict():
df = get_model_input_df()
model_res_dict = {}
# fixed effects meta analyses (lnVR and lnCVR)
for model in ['fema', 'rema'
]: # lnVR, # random effects meta analyses (lnVR and lnCVR)
stan_model = compile_model(os.path.join(stan_model_path,
f'{model}.stan'),
model_name=model)
for effect_statistic in ['lnVR', 'lnCVR']:
data_dict = get_data_dict(df, effect_statistic)
fit = stan_model.sampling(data=data_dict,
iter=4000,
warmup=1000,
chains=3,
control={'adapt_delta': 0.99},
check_hmc_diagnostics=True,
seed=1)
data = az.from_pystan(
posterior=fit,
posterior_predictive=['Y_pred'],
observed_data=['Y'],
log_likelihood='log_lik',
)
model_res_dict[f'{model}_{effect_statistic}'] = data
model = 'remr'
stan_model = compile_model(os.path.join(stan_model_path, f'{model}.stan'),
model_name=model)
effect_statistic = 'lnVR'
data_dict = get_data_dict(df, effect_statistic)
fit = stan_model.sampling(data=data_dict,
iter=4000,
warmup=1000,
chains=3,
control={'adapt_delta': 0.99},
check_hmc_diagnostics=True,
seed=1)
pystan.check_hmc_diagnostics(fit)
data = az.from_pystan(
posterior=fit,
posterior_predictive=['Y_pred'],
observed_data=['Y_meas', 'X_meas'],
log_likelihood='log_lik',
)
model_res_dict[f'{model}_{effect_statistic}'] = data
return model_res_dict
def fit_models(stan_data, stan_models, **hmc_args):
fits = []
az_fits = {k: None for k in fit_models}
for fit_model in stan_models:
print('Fitting %s model' % fit_model)
model = pkl.load(
open(CODE_DIR + 'assignment_errors/%s_model.stan' % fit_model,
'rb'))
fit = model.sampling(data=stan_data, **hmc_args)
model_params = {
'observed_data': 'y',
'log_likelihood': {
'y': 'log_lik'
},
'posterior_predictive': 'err_hat'
}
fit_az = az.from_pystan(posterior=fit, **model_params)
fits.append(fit)
az_fits[fit_model] = fit_az
comps = az.compare(az_fits)
return comps, fits
def get_varying_intercept_model_results():
# read in Cipriani data
df = get_model_input_df()
data_dict = {
'N': df.shape[0],
'Y_meas': df['lnSD'].values,
'X_meas': df['lnMean'].values,
'SD_Y': np.sqrt(df['var_lnSD'].values),
'SD_X': np.sqrt(df['var_lnMean'].values),
'K': len(df.scale.unique()),
'scale_group': df.scale_rank.values
}
varying_intercept_stan_model = compile_model(
os.path.join(stan_model_path, 'varying_intercept_regression.stan'),
model_name='varying_intercept_regression')
fit = varying_intercept_stan_model.sampling(data=data_dict,
iter=4000,
warmup=1000,
chains=3,
control={'adapt_delta': 0.99},
check_hmc_diagnostics=True,
seed=1)
pystan.check_hmc_diagnostics(fit)
data = az.from_pystan(
posterior=fit,
posterior_predictive=['Y_pred'],
observed_data=['X_meas', 'Y_meas'],
log_likelihood='log_lik',
)
return data
def print_fit(*args: TaskModel, ic: str = 'looic') -> pd.DataFrame:
"""Print model-fits (mean LOOIC or WAIC values) of hbayesdm models.
Parameters
----------
args
Output instances of running hbayesdm model functions.
ic
Information criterion (defaults to 'looic').
Returns
-------
pd.DataFrame
Model-fit info per each hbayesdm output given as argument(s).
"""
ic_options = ('looic', 'waic')
if ic not in ic_options:
raise RuntimeError('Information Criterion (ic) must be one of ' +
repr(ic_options))
dataset_dict = {
model_data.model: az.from_pystan(model_data.fit,
log_likelihood='log_lik')
for model_data in args
}
ic = 'loo' if ic == 'looic' else 'waic'
return az.compare(dataset_dict=dataset_dict, ic=ic)
def test_empty_parameter(self):
if pystan_version() == 2:
model_code = """
parameters {
real y;
vector[3] x;
vector[0] a;
vector[2] z;
}
model {
y ~ normal(0,1);
}
"""
from pystan import StanModel # pylint: disable=import-error
model = StanModel(model_code=model_code)
fit = model.sampling(iter=10,
chains=2,
check_hmc_diagnostics=False)
posterior = from_pystan(posterior=fit)
test_dict = {
"posterior": ["y", "x", "z"],
"sample_stats": ["diverging"]
}
fails = check_multiple_attrs(test_dict, posterior)
assert not fails
def get_waic_and_loo(fit):
"""Compute WAIC and LOO from a fit instance"""
idata = az.from_pystan(fit, log_likelihood="llx")
result = {}
result.update(dict(az.loo(idata, scale='deviance')))
result.update(dict(az.waic(idata, scale='deviance')))
return result
def test_index_order(self, data, eight_schools_params):
"""Test 0-indexed data."""
# Skip test if pystan not installed
pystan = importorskip("pystan") # pylint: disable=import-error
fit = data.model.sampling(data=eight_schools_params)
if pystan.__version__ >= "2.18":
# make 1-indexed to 0-indexed
for holder in fit.sim["samples"]:
new_chains = OrderedDict()
for i, (key, values) in enumerate(holder.chains.items()):
if "[" in key:
name, *shape = key.replace("]", "").split("[")
shape = [
str(int(item) - 1) for items in shape
for item in items.split(",")
]
key = name + "[{}]".format(",".join(shape))
new_chains[key] = np.full_like(values, fill_value=float(i))
setattr(holder, "chains", new_chains)
fit.sim["fnames_oi"] = list(fit.sim["samples"][0].chains.keys())
idata = from_pystan(posterior=fit)
assert idata is not None
for j, fpar in enumerate(fit.sim["fnames_oi"]):
if fpar == "lp__":
continue
par, *shape = fpar.replace("]", "").split("[")
assert hasattr(idata.posterior, par)
if shape:
shape = [slice(None), slice(None)] + list(map(int, shape))
assert idata.posterior[par][tuple(
shape)].values.mean() == float(j)
else:
assert idata.posterior[par].values.mean() == float(j)
def test_invalid_fit(self, data):
if pystan_version() == 2:
model = data.model
model_data = {
"J": 8,
"y": np.array([28.0, 8.0, -3.0, 7.0, -1.0, 1.0, 18.0, 12.0]),
"sigma": np.array([15.0, 10.0, 16.0, 11.0, 9.0, 11.0, 10.0, 18.0]),
}
fit_test_grad = model.sampling(
data=model_data, test_grad=True, check_hmc_diagnostics=False
)
with pytest.raises(AttributeError):
_ = from_pystan(posterior=fit_test_grad)
fit = model.sampling(data=model_data, iter=100, chains=1, check_hmc_diagnostics=False)
del fit.sim["samples"]
with pytest.raises(AttributeError):
_ = from_pystan(posterior=fit)
def __init__(self, data, dataset, ci=95.):
if az is None:
raise ValueError("ArviZ package must be installed in order to work"
" with the BayesianMetaRegressionResults class.")
if data.__class__.__name__ == 'StanFit4Model':
data = az.from_pystan(data)
self.data = data
self.dataset = dataset
self.ci = ci
def fit_bhv_model(data,
model_path=bmp,
targ_field='LABthetaTarget',
dist_field='LABthetaDist',
resp_field='LABthetaResp',
prior_dict=None,
stan_iters=2000,
stan_chains=4,
arviz=mixture_arviz,
adapt_delta=.9,
diagnostics=True,
**stan_params):
if prior_dict is None:
prior_dict = default_prior_dict
targs_is = data[targ_field]
session_list = np.array(data[['animal', 'date']])
mapping_list = []
session_nums = np.array([], dtype=int)
for i, x in enumerate(targs_is):
sess = np.ones(len(x), dtype=int) * (i + 1)
session_nums = np.concatenate((session_nums, sess))
indices = x.index
sess_info0 = (str(session_list[i, 0]), ) * len(x)
sess_info1 = (str(session_list[i, 1]), ) * len(x)
mapping_list = mapping_list + list(zip(indices, sess_info0,
sess_info1))
mapping_dict = {i: mapping_list[i] for i in range(len(session_nums))}
targs = np.concatenate(targs_is, axis=0)
dists = np.concatenate(data[dist_field], axis=0)
resps = np.concatenate(data[resp_field], axis=0)
errs = u.normalize_periodic_range(targs - resps)
dist_errs = u.normalize_periodic_range(dists - resps)
dists_per = u.normalize_periodic_range(dists - targs)
stan_data = dict(T=dist_errs.shape[0],
S=len(targs_is),
err=errs,
dist_err=dist_errs,
run_ind=session_nums,
dist_loc=dists_per,
**prior_dict)
control = {
'adapt_delta': stan_params.pop('adapt_delta', .8),
'max_treedepth': stan_params.pop('max_treedepth', 10)
}
sm = pickle.load(open(model_path, 'rb'))
fit = sm.sampling(data=stan_data,
iter=stan_iters,
chains=stan_chains,
control=control,
**stan_params)
if diagnostics:
diag = ps.diagnostics.check_hmc_diagnostics(fit)
else:
diag = None
fit_av = az.from_pystan(posterior=fit, **arviz)
return fit, diag, fit_av, stan_data, mapping_dict
def get_inference_data4(self, data):
"""multiple vars as lists."""
return from_pystan(
posterior=data.obj,
posterior_predictive=None,
prior=data.obj,
prior_predictive=None,
observed_data="y",
coords=None,
dims=None,
)
def get_inference_data4(self, data):
"""minimal input."""
return from_pystan(
posterior=data.obj,
posterior_predictive=None,
prior=data.obj,
prior_predictive=None,
coords=None,
dims=None,
posterior_model=data.model,
prior_model=data.model,
)
def get_inference_data3(self, data, eight_schools_params):
"""multiple vars as lists."""
return from_pystan(
posterior=data.obj,
posterior_predictive=["y_hat", "log_lik"],
prior=data.obj,
prior_predictive=["y_hat", "log_lik"],
observed_data="y",
coords={"school": np.arange(eight_schools_params["J"])},
dims={"theta": ["school"], "y": ["school"], "y_hat": ["school"], "eta": ["school"]},
posterior_model=data.model,
prior_model=data.model,
)
def get_inference_data(self):
return from_pystan(fit=self.obj,
posterior_predictive='y_hat',
observed_data=['y'],
log_likelihood='log_lik',
coords={'school': np.arange(self.data['J'])},
dims={
'theta': ['school'],
'y': ['school'],
'log_lik': ['school'],
'y_hat': ['school'],
'theta_tilde': ['school']
})
def create_diagnostic_plots(idx,pdf_filename,fit,diag_pars,niter,nchain):
# Converting the Stan FIT object to Arviz InfereceData
samples = fit.extract(permuted=True) # Extracting parameter samples
data = az.from_pystan(fit)
tmp = data.posterior
var_names = list(tmp.data_vars)
# Filtering the list of parameters to plot
unwanted = {'losvd','spec','conv_spec','poly','bestfit','losvd_','losvd_mod','spec_pred','log_likelihood'}
vars_main = [e for e in var_names if e not in unwanted]
# Reading diagnostic parameters
accept_stat, stepsize, treedepth = np.zeros((niter,nchain)), np.zeros((niter,nchain)) , np.zeros((niter,nchain))
n_leapfrog, divergent, energy = np.zeros((niter,nchain)), np.zeros((niter,nchain)) , np.zeros((niter,nchain))
for j in range(nchain):
accept_stat[:,j] = diag_pars[j]['accept_stat__']
stepsize[:,j] = diag_pars[j]['stepsize__']
treedepth[:,j] = diag_pars[j]['treedepth__']
n_leapfrog[:,j] = diag_pars[j]['n_leapfrog__']
divergent[:,j] = diag_pars[j]['divergent__']
energy[:,j] = diag_pars[j]['energy__']
# Creating the plot in multiple PDF papges
pdf_pages = PdfPages(pdf_filename)
print(" - Sampler params")
plot_sampler_params(idx,accept_stat,stepsize,treedepth,n_leapfrog,divergent,energy)
pdf_pages.savefig()
print(" - Chains")
plot_chains(samples,vars_main)
pdf_pages.savefig()
# print(" - Trace plot [Main params]")
# az.plot_trace(data, var_names=vars_main)
# pdf_pages.savefig()
# print(" - Trace plot [LOSVD]")
# az.plot_trace(data, var_names=['losvd'])
# pdf_pages.savefig()
print(" - Pair plot")
az.plot_pair(data, var_names=vars_main, divergences=True, kind='kde', fill_last=False)
pdf_pages.savefig()
print(" - Autocorr plot")
az.plot_autocorr(data, var_names=vars_main)
pdf_pages.savefig()
print(" - Energy plot")
az.plot_energy(data)
pdf_pages.savefig()
pdf_pages.close()
return
def get_inference_data5(self, data):
"""minimal input."""
return from_pystan(
posterior=data.obj,
posterior_predictive=None,
prior=data.obj,
prior_predictive=None,
coords=None,
dims=None,
posterior_model=data.model,
log_likelihood=False,
prior_model=data.model,
save_warmup=pystan_version() == 2,
)
def get_inference_data(self):
return from_pystan(
fit=self.obj,
posterior_predictive="y_hat",
observed_data=["y"],
log_likelihood="log_lik",
coords={"school": np.arange(self.data["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"log_lik": ["school"],
"y_hat": ["school"],
"theta_tilde": ["school"],
},
)
def fit(self, y1, y2, w1=None, w2=None):
self.mcmc_ = get_mcmc(self.model, y1, y2, w1, w2, **self.kwargs)
self.data_ = az.from_pystan(
posterior=self.mcmc_,
posterior_predictive=['y1_pred', 'y2_pred'],
observed_data=['y1', 'y2'],
log_likelihood='log_lik',
coords={
'group_mu': ['Group 1', 'Group 2'],
'group_sigma': ['Group 1', 'Group 2']
},
dims={
'mu': ['group_mu'],
'sigma': ['group_sigma']
})
def get_inference_data2(self):
# dictionary
observed_data = {'y_hat': self.data['y']}
# ndarray
log_likelihood = self.obj.extract('log_lik', permuted=False)['log_lik']
return from_pystan(fit=self.obj,
posterior_predictive='y_hat',
observed_data=observed_data,
log_likelihood=log_likelihood,
coords={'school': np.arange(self.data['J'])},
dims={
'theta': ['school'],
'y': ['school'],
'log_lik': ['school'],
'y_hat': ['school'],
'theta_tilde': ['school']
})
def get_subgroup_models():
df = get_formatted_data()
# drug class subgroup analysis
model_res_dict = {}
for drug_class in DRUG_CLASSES:
study_ids = df.query(f'drug_class == "{drug_class}"').study_id.unique()
df_sub = df[(df.study_id.isin(study_ids))
& (df.drug_class.isin([drug_class, 'placebo']))].copy()
placebo_controlled_study_ids = set(df_sub.query('is_active == 1')['study_id']) \
.intersection(df_sub.query('is_active == 0')['study_id'])
df_sub = df_sub[df_sub.study_id.isin(placebo_controlled_study_ids)]
for column in ['study_id', 'scale', 'drug_class']:
df_sub = add_rank_column(df_sub, column)
df_sub = aggregate_treatment_arms(df_sub)
df_sub = get_variability_effect_sizes(df_sub)
model = 'remr'
stan_model = compile_model(os.path.join(stan_model_path,
f'{model}.stan'),
model_name=model)
data_dict = get_data_dict(df_sub, 'lnVR')
fit = stan_model.sampling(data=data_dict,
iter=4000,
warmup=1000,
chains=3,
control={'adapt_delta': 0.99},
check_hmc_diagnostics=True,
seed=1)
pystan.check_hmc_diagnostics(fit)
data = az.from_pystan(
posterior=fit,
posterior_predictive=['Y_pred'],
observed_data=['Y_meas', 'X_meas'],
log_likelihood='log_lik',
)
model_res_dict[drug_class] = data
return model_res_dict
def get_inference_data3(self, data, eight_schools_params):
"""log_likelihood as a ndarray."""
# ndarray
log_likelihood = pystan_extract_normal(data.obj, "log_lik")["log_lik"]
return from_pystan(
fit=data.obj,
posterior_predictive=["y_hat"],
observed_data=["y"],
log_likelihood=log_likelihood,
coords={"school": np.arange(eight_schools_params["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"log_lik": ["school"],
"y_hat": ["school"],
"theta_tilde": ["school"],
},
)
def get_inference_data(self, data, eight_school_params):
"""vars as str."""
return from_pystan(
posterior=data.obj,
posterior_predictive="y_hat",
prior=data.obj,
prior_predictive="y_hat",
observed_data="y",
log_likelihood="log_lik",
coords={"school": np.arange(eight_school_params["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"log_lik": ["school"],
"y_hat": ["school"],
"theta_tilde": ["school"],
},
)
def get_inference_data2(self):
# dictionary
observed_data = {"y_hat": self.data["y"]}
# ndarray
log_likelihood = self.obj.extract("log_lik", permuted=False)["log_lik"]
return from_pystan(
fit=self.obj,
posterior_predictive="y_hat",
observed_data=observed_data,
log_likelihood=log_likelihood,
coords={"school": np.arange(self.data["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"log_lik": ["school"],
"y_hat": ["school"],
"theta_tilde": ["school"],
},
)
def get_inference_data(self, data, eight_school_params):
"""log_likelihood as a var."""
prior = pystan_extract_unpermuted(data.obj)
prior = {"theta_test": prior["theta"]}
return from_pystan(
fit=data.obj,
prior=prior,
posterior_predictive="y_hat",
observed_data=["y"],
log_likelihood="log_lik",
coords={"school": np.arange(eight_school_params["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"log_lik": ["school"],
"y_hat": ["school"],
"theta_tilde": ["school"],
},
)
def run(self, samples=1000, chains=1, **kwargs): # pylint: disable=arguments-differ
"""
Run the Stan sampler.
Parameters
----------
samples : int
Number of samples to obtain (in each chain).
chains : int
Number of chains to use.
kwargs : dict
Optional keyword arguments passed onto the PyStan StanModel.sampling() call.
Returns:
----------
An ArviZ InferenceData instance.
"""
self.fit = self.stan_model.sampling(data=self.X, iter=samples, chains=chains, **kwargs)
return from_pystan(self.fit)
def test_empty_parameter(self):
if pystan_version() == 2:
model_code = """
parameters {
real y;
vector[0] z;
}
model {
y ~ normal(0,1);
}
"""
from pystan import StanModel
model = StanModel(model_code=model_code)
fit = model.sampling(iter=10, chains=2, check_hmc_diagnostics=False)
posterior = from_pystan(posterior=fit)
assert hasattr(posterior, "posterior")
assert hasattr(posterior.posterior, "y")
assert not hasattr(posterior.posterior, "z")
def get_baseline_severity_model():
df = prepare_data()
effect_statistic = 'lnVR'
data_dict = {
'N': len(df.study_id.unique()),
'Y_meas': df.groupby(['study_id']).agg({effect_statistic: 'first'}).reset_index()[effect_statistic].values,
'X_meas': df.groupby(['study_id']).agg({'lnRR': 'first'}).reset_index()['lnRR'].values,
'SD_Y': np.sqrt(df.groupby(['study_id']).agg(
{f'var_{effect_statistic}': 'first'}).reset_index()[f'var_{effect_statistic}'].values),
'SD_X': np.sqrt(df.groupby(['study_id']).agg(
{'var_lnRR': 'first'}).reset_index()['var_lnRR'].values),
'X0': df.groupby(['study_id']).apply(
lambda x: np.sum(x['baseline'] * x['N']) / np.sum(x['N'])
).reset_index()[0].values,
'run_estimation': 1
}
stan_model = compile_model(
os.path.join(stan_model_path, 'remr_bs.stan'),
model_name='remr_bs'
)
fit = stan_model.sampling(
data=data_dict,
iter=4000,
warmup=1000,
chains=3,
control={'adapt_delta': 0.99},
check_hmc_diagnostics=True,
seed=1
)
pystan.check_hmc_diagnostics(fit)
data = az.from_pystan(
posterior=fit,
posterior_predictive=['Y_pred'],
observed_data=['Y_meas', 'X_meas', 'X0'],
log_likelihood='log_lik',
)
return data
def get_simulation_results():
data_dict = {
'N': 1000,
'rho': -0.4,
'sd_te': 6.5,
'sd_m': 0.001,
'lambda': 0.2,
'theta': 0.9
}
simulation_stan_model = compile_model(os.path.join(stan_model_path,
'simulation.stan'),
model_name='simulation')
fit = simulation_stan_model.sampling(data=data_dict,
warmup=500,
iter=2500,
chains=2,
check_hmc_diagnostics=True,
seed=1)
pystan.check_hmc_diagnostics(fit)
data = az.from_pystan(posterior=fit)
return data
def get_inference_data3(self, data, eight_schools_params):
"""multiple vars as lists."""
return from_pystan(
posterior=data.obj,
posterior_predictive=["y_hat", "log_lik"], # wrong, but fine for testing
predictions=["y_hat", "log_lik"], # wrong, but fine for testing
prior=data.obj,
prior_predictive=["y_hat", "log_lik"], # wrong, but fine for testing
constant_data=["sigma", "y"], # wrong, but fine for testing
predictions_constant_data=["sigma", "y"], # wrong, but fine for testing
coords={"school": np.arange(eight_schools_params["J"])},
dims={
"theta": ["school"],
"y": ["school"],
"sigma": ["school"],
"y_hat": ["school"],
"eta": ["school"],
},
posterior_model=data.model,
prior_model=data.model,
)
def run_model(rankings,
survey_head_to_heads,
stan_file=STAN_FILE,
model_config='combined'):
n_episode_contestant = rankings.groupby(
'episode_id')['contestant_id'].nunique()
episode_rank_counts = (rankings.groupby(
['episode_id', 'rank']).size().unstack().fillna(0).astype(int))
contestants = rankings.groupby('contestant_id').first()
contestants['id_stan'] = range(1, len(contestants) + 1)
rankings = rankings.join(contestants['id_stan'], on='contestant_id')
survey_head_to_heads = (survey_head_to_heads.join(
contestants['id_stan'].rename('id_stan_own'),
on='own').join(contestants['id_stan'].rename('id_stan_opp'), on='opp'))
input_data = {
'N': len(rankings),
'K': len(PREDICTORS),
'C': len(contestants),
'E': rankings['episode_id'].nunique(),
'X': contestants[PREDICTORS].fillna(0).values,
'N_episode_contestant': n_episode_contestant.values,
'episode_rank': rankings['rank'].values,
'contestant': rankings['id_stan'].values,
'N_survey': len(survey_head_to_heads),
'survey_contestant': survey_head_to_heads['id_stan_own'].values,
'survey_opponent': survey_head_to_heads['id_stan_opp'].values,
'survey_count': survey_head_to_heads['count'].values,
'survey_wins': survey_head_to_heads['wins'].values
}
model = StanModel_cache(file=stan_file)
fit = model.sampling(data={**input_data, **model_config})
return arviz.from_pystan(fit,
coords={
'contestant': contestants.index,
'predictor': PREDICTORS
},
dims={
'ability': ['contestant'],
'beta': ['predictor']
})
