def test_extract_dag():
import pymc3 as pm
model = pm.Model()
# test empty model
dag = get_dag(model)
assert len(dag) == 0
# 3 basic variables
with model:
left_slope = pm.Normal("left_slope", mu=0, sd=0.5)
right_slope = pm.Normal("right_slope", mu=0, sd=0.5)
switchpoint = pm.Normal("switchpoint", mu=0, sd=100)
dag = get_dag(model)
assert len(dag) == 3
with model:
y_obs = pm.Normal(
"y_obs", mu=switchpoint, sd=left_slope, observed=np.ones((10,))
)
rescaled = pm.Deterministic("rescaled", np.log(np.abs(left_slope)))
penalty = pm.Potential("potential", right_slope)
# check all six nodes are accounted for
dag = get_dag(model)
assert len(dag) == 6
# check parents
assert "left_slope" in dag["y_obs"]["parents"]
assert "switchpoint" in dag["y_obs"]["parents"]
assert "left_slope" in dag["rescaled"]["parents"]
assert "right_slope" not in dag["y_obs"]["parents"]
# check variable types
assert dag["left_slope"]["type"] == "free"
assert dag["y_obs"]["type"] == "observed"
assert dag["potential"]["type"] == "potential"
assert dag["rescaled"]["type"] == "deterministic"
assert dag["y_obs"]["distribution"]["type"] == "Normal"
assert dag["left_slope"]["distribution"]["type"] == "Normal"
def capture_inference(model, samples=1000, chains=4, predictive=500):
with model:
# Get posterior trace, prior trace, posterior predictive samples, and the DAG
trace = pm.sample(samples=samples, chains=chains)
prior = pm.sample_prior_predictive(samples=samples)
posterior_predictive = pm.sample_posterior_predictive(trace)
dag = get_dag(model)
# will also capture all the sampler statistics
data = az.from_pymc3(trace=trace,
prior=prior,
posterior_predictive=posterior_predictive)
# insert dag into sampler stat attributes
data.sample_stats.attrs["graph"] = dag
return data
fileName='golf_geometry_PyMC3'
samples=2000
chains=2
tune=1000
geometry_model=pm.Model(coords=coords)
with geometry_model:
#to store the n-parameter of Binomial dist
#in the constant group of ArviZ InferenceData
#You should always call it n for imd to retrieve it
n = pm.Data('n', data.tries)
sigma_angle = pm.HalfNormal('sigma_angle')
p_goes_in = pm.Deterministic('p_goes_in', 2 * Phi(tt.arcsin((CUP_RADIUS - BALL_RADIUS) / data.distance) / sigma_angle) - 1, dims='distance')
successes = pm.Binomial('successes', n=n, p=p_goes_in, observed=data.successes, dims='distance')
#inference
trace_g = pm.sample(draws=samples, chains=chains, tune=tune)
prior_g= pm.sample_prior_predictive(samples=samples)
posterior_predictive_g = pm.sample_posterior_predictive(trace_g,samples=samples)
## STEP 1
# will also capture all the sampler statistics
data_g = az.from_pymc3(trace=trace_g, prior=prior_g, posterior_predictive=posterior_predictive_g)
## STEP 2
#dag
dag_g = get_dag(geometry_model)
# insert dag into sampler stat attributes
data_g.sample_stats.attrs["graph"] = str(dag_g)
## STEP 3
# save data
arviz_to_json(data_g, fileName+'.npz')
sigma = np.array([15., 10., 16., 11., 9., 11., 10., 18.])
#model-inference
coords_c = {"school": ["A","B","C","D","E","F","G","H"]}
fileName_c="eight_schools_centered"
samples=4000
chains=2
tune=1000
with pm.Model(coords=coords_c) as centered_eight:
mu = pm.Normal('mu', mu=0, sigma=5)
tau = pm.HalfCauchy('tau', beta=5)
theta = pm.Normal('theta', mu=mu, sigma=tau, dims='school')
y = pm.Normal('y', mu=theta, sigma=sigma, observed=obs, dims='school')
#inference
trace_c = pm.sample(samples, chains=chains, tune=tune, random_seed=SEED)
prior_c= pm.sample_prior_predictive(samples=samples)
posterior_predictive_c = pm.sample_posterior_predictive(trace_c, samples=samples)
## STEP 1
# will also capture all the sampler statistics
data_c = az.from_pymc3(trace = trace_c, prior = prior_c, posterior_predictive = posterior_predictive_c)
## STEP 2
#dag
dag_c = get_dag(centered_eight)
# insert dag into sampler stat attributes
data_c.sample_stats.attrs["graph"] = str(dag_c)
## STEP 3
# save data
arviz_to_json(data_c, fileName_c+'.npz')
a = pm.Normal("a", mu=mu_a, sd=sigma_a, dims="driver")
b = pm.Normal("b", mu=mu_b, sd=sigma_b, dims="driver")
sigma = pm.HalfNormal("sigma", sd=sigma_sigma, dims="driver")
y_pred = pm.Normal('y_pred',
mu=a[driver_idx] + b[driver_idx] * day_idx,
sd=sigma[driver_idx],
observed=reactions.Reaction,
dims="driver_idx_day")
## inference
trace_hi = pm.sample(draws=samples, chains=chains, tune=tune)
prior_hi = pm.sample_prior_predictive(samples=samples)
posterior_predictive_hi = pm.sample_posterior_predictive(trace_hi,
samples=samples)
## STEP 1
## export inference results in ArviZ InferenceData obj
## will also capture all the sampler statistics
data_hi = az.from_pymc3(trace=trace_hi,
prior=prior_hi,
posterior_predictive=posterior_predictive_hi)
## STEP 2
## extract dag
dag_hi = get_dag(hierarchical_model)
## insert dag into sampler stat attributes
data_hi.sample_stats.attrs["graph"] = str(dag_hi)
## STEP 3
## save data
fileName_hi = "reaction_times_hierarchical"
arviz_to_json(data_hi, fileName_hi + '.npz')
#model-inference
coords = {"school": ["A","B","C","D","E","F","G","H"]}
samples=5000
chains=2
tune=1000
fileName="eight_schools_non_centered"
with pm.Model(coords=coords) as NonCentered_eight:
mu = pm.Normal('mu', mu=0, sigma=5)
tau = pm.HalfCauchy('tau', beta=5)
theta_tilde = pm.Normal('theta_t', mu=0, sigma=1, dims='school')
theta = pm.Deterministic('theta', mu + tau * theta_tilde, dims='school')
y = pm.Normal('y', mu=theta, sigma=sigma, observed=obs, dims='school')
#inference
trace_nc = pm.sample(samples, chains=chains, tune=tune, random_seed=SEED, target_accept=.90)
prior_nc= pm.sample_prior_predictive(samples=samples)
posterior_predictive_nc = pm.sample_posterior_predictive(trace_nc,samples=samples)
## STEP 1
# will also capture all the sampler statistics
data_nc = az.from_pymc3(trace = trace_nc, prior = prior_nc, posterior_predictive = posterior_predictive_nc)
## STEP 2
#dag
dag_nc = get_dag(NonCentered_eight)
# insert dag into sampler stat attributes
data_nc.sample_stats.attrs["graph"] = str(dag_nc)
## STEP 3
# save data
arviz_to_json(data_nc, fileName+'.npz')
trace = pm.sample(samples=samples, chains=chains)
prior = pm.sample_prior_predictive(samples=samples)
posterior_predictive = pm.sample_posterior_predictive(trace)
dag = get_dag(model)
# will also capture all the sampler statistics
data = az.from_pymc3(trace=trace,
prior=prior,
posterior_predictive=posterior_predictive)
# insert dag into sampler stat attributes
data.sample_stats.attrs["graph"] = dag
return data
if __name__ == "__main__":
# generate a single switchpoint model
poverty = load_data()
model = define_model(poverty)
dag = get_dag(model)
data = capture_inference(model)
arviz_to_json(data, "switchpoint.npz")
# generate multiple models
models = {
"centered": az.load_arviz_data("centered_eight"),
"noncentered": az.load_arviz_data("non_centered_eight"),
}
multi_arviz_to_json(models, "multimodel.zip")
dates=returns.index.strftime("%Y/%m/%d").tolist()
#model-inference
fileName='stochastic_volatility_PyMC3'
samples=2000
tune=2000
chains=2
coords = {"date": dates}
with pm.Model(coords=coords) as model:
step_size = pm.Exponential('step_size', 10)
volatility = pm.GaussianRandomWalk('volatility', sigma=step_size, dims='date')
nu = pm.Exponential('nu', 0.1)
returns = pm.StudentT('returns',nu=nu,lam=np.exp(-2*volatility) ,observed=data["change"], dims='date')
#inference
trace = pm.sample(draws=samples, chains=chains, tune=tune)
prior = pm.sample_prior_predictive(samples=samples)
posterior_predictive = pm.sample_posterior_predictive(trace,samples=samples)
## STEP 1
# will also capture all the sampler statistics
data = az.from_pymc3(trace=trace, prior=prior, posterior_predictive=posterior_predictive)
## STEP 2
#dag
dag = get_dag(stochastic_vol_model)
# insert dag into sampler stat attributes
data.sample_stats.attrs["graph"] = str(dag)
## STEP 3
# save data
arviz_to_json(data, fileName+'.npz')
#model-inference
fileName='coal_mining_disasters_PyMC3'
samples=10000
tune=10000
chains=2
coords = {"year": years}
with pm.Model(coords=coords) as disaster_model:
switchpoint = pm.DiscreteUniform('switchpoint', lower=years.min(), upper=years.max(), testval=1900)
early_rate = pm.Exponential('early_rate', 1)
late_rate = pm.Exponential('late_rate', 1)
rate = pm.math.switch(switchpoint >= years, early_rate, late_rate)
disasters = pm.Poisson('disasters', rate, observed=disaster_data, dims='year')
#inference
trace = pm.sample(samples, chains=chains, tune=tune)
prior = pm.sample_prior_predictive(samples=samples)
posterior_predictive = pm.sample_posterior_predictive(trace,samples=samples)
## STEP 1
# will also capture all the sampler statistics
data = az.from_pymc3(trace=trace, prior=prior, posterior_predictive=posterior_predictive)
## STEP 2
# extract dag
dag = get_dag(disaster_model)
# insert dag into sampler stat attributes
data.sample_stats.attrs["graph"] = str(dag)
## STEP 3
# save data
arviz_to_json(data, fileName+'.npz')
a = pm.Normal("a", mu=100, sd=250)
b = pm.Normal("b", mu=10, sd=250)
sigma = pm.HalfNormal("sigma", sd=200)
y_pred = pm.Normal('y_pred',
mu=a + b * day_idx,
sd=sigma,
observed=reactions.Reaction,
dims="driver_idx_day")
## inference
trace_p = pm.sample(samples, chains=chains, tune=tune)
prior_p = pm.sample_prior_predictive(samples=samples)
posterior_predictive_p = pm.sample_posterior_predictive(trace_p,
samples=samples)
## STEP 1
## export inference results in ArviZ InferenceData obj
## will also capture all the sampler statistics
data_p = az.from_pymc3(trace=trace_p,
prior=prior_p,
posterior_predictive=posterior_predictive_p)
## STEP 2
## extract dag
dag_p = get_dag(fullyPooled_model)
## insert dag into sampler stat attributes
data_p.sample_stats.attrs["graph"] = str(dag_p)
## STEP 3
## save data
fileName_p = "reaction_times_pooled"
arviz_to_json(data_p, fileName_p + '.npz')
chains=2
tune=1000
simple_model=pm.Model(coords=coords)
with simple_model:
#to store the n-parameter of Binomial dist
#in the constant group of ArviZ InferenceData
#You should always call it n for imd to retrieve it
n = pm.Data('n', data.tries)
a = pm.Normal('a')
b = pm.Normal('b')
p_goes_in = pm.Deterministic('p_goes_in', pm.math.invlogit(a * data.distance + b), dims='distance')
successes = pm.Binomial('successes', n=n, p=p_goes_in, observed=data.successes, dims='distance')
#inference
# Get posterior trace, prior trace, posterior predictive samples, and the DAG
trace = pm.sample(draws=samples, chains=chains, tune=tune)
prior= pm.sample_prior_predictive(samples=samples)
posterior_predictive = pm.sample_posterior_predictive(trace,samples=samples)
## STEP 1
# will also capture all the sampler statistics
data_s = az.from_pymc3(trace=trace, prior=prior, posterior_predictive=posterior_predictive)
## STEP 2
#dag
dag = get_dag(simple_model)
# insert dag into sampler stat attributes
data_s.sample_stats.attrs["graph"] = str(dag)
## STEP 3
# save data
arviz_to_json(data_s, fileName+'.npz')