def perform_SALib_sobol(model, model_output, output_name, subplot):
# Returns a dict with a problem specificatin as required by SALib
problem = get_SALib_problem(model.uncertainties)
Si = sobol.analyze(problem,
model_output,
calc_second_order=True,
print_to_console=False)
# If the total-order indices are substantially larger than the first-order indices,
# then there is likely higher-order interactions occurring. We can look at the second-order indices
# to see these higher-order interactions
scores_filtered = {k: Si[k] for k in ['ST', 'ST_conf', 'S1', 'S1_conf']}
Si_df = pd.DataFrame(scores_filtered, index=problem['names'])
Si_df.sort_values(by='ST', ascending=False)
sns.set_style('white')
indices = Si_df[['S1', 'ST']]
err = Si_df[['S1_conf', 'ST_conf']]
indices.plot.bar(yerr=err.values.T, ax=subplot)
print("\n" + output_name)
print("")
print(Si_df)
def test_get_salib_problem(self):
uncertainties = [RealParameter("a", 0, 10), RealParameter("b", 0, 5)]
problem = get_SALib_problem(uncertainties)
self.assertEqual(2, problem['num_vars'])
self.assertEqual(['a', 'b'], problem['names'])
self.assertEqual((0, 10), problem['bounds'][0])
self.assertEqual((0, 5), problem['bounds'][1])
def test_get_salib_problem(self):
uncertainties = [RealParameter("a", 0, 10),
RealParameter("b", 0, 5)]
problem = get_SALib_problem(uncertainties)
self.assertEqual(2, problem['num_vars'])
self.assertEqual(['a', 'b'], problem['names'])
self.assertEqual((0, 10), problem['bounds'][0])
self.assertEqual((0, 5), problem['bounds'][1])
def plot_sobol(model, sa_results, prob):
if prob == "uncertainties":
problem = get_SALib_problem(model.uncertainties)
elif prob == "levers":
problem = get_SALib_problem(model.levers)
outcomes = sa_results[1]
n_outcome = len(outcomes.keys())
ncols = 4
nrows = math.ceil(n_outcome / ncols)
fig, axes = plt.subplots(figsize=(ncols * 4, nrows * 4),
ncols=ncols,
nrows=nrows)
locs = list(itertools.product(range(nrows), range(ncols)))
for i, key in enumerate(outcomes.keys()):
loc = locs[i]
ax = axes[loc]
outcome = outcomes[key]
Si = sobol.analyze(problem,
outcome,
calc_second_order=True,
print_to_console=False)
Si_filter = {k: Si[k] for k in ['ST', 'ST_conf', 'S1', 'S1_conf']}
Si_df = pd.DataFrame(Si_filter, index=problem['names'])
Si_df = Si_df.T.stack().reset_index()
Si_df.columns = ["S", "Uncertainty", "Sensitivity Value"]
Si_df["Uncertainty"] = Si_df["Uncertainty"].apply(
lambda x: x[:9]).values
sns.barplot(x="Uncertainty",
y="Sensitivity Value",
data=Si_df,
hue="S",
ax=ax)
ax.set_xlabel("")
ax.set_xticklabels(labels=Si_df["Uncertainty"].unique(), rotation=90)
ax.set_title(key)
ax.grid()
fig.tight_layout()
]
from ema_workbench import (SequentialEvaluator, ema_logging,
perform_experiments)
ema_logging.log_to_stderr(ema_logging.INFO)
from SALib.analyze import sobol
from ema_workbench.em_framework.salib_samplers import get_SALib_problem
with SequentialEvaluator(model) as evaluator:
sa_results = evaluator.perform_experiments(scenarios=1,
uncertainty_sampling='sobol')
experiments, outcomes = sa_results
problem = get_SALib_problem(model.uncertainties)
Si = sobol.analyze(problem,
outcomes['max_P'],
calc_second_order=True,
print_to_console=False)
scores_filtered = {k: Si[k] for k in ['ST', 'ST_conf', 'S1', 'S1_conf']}
Si_df = pd.DataFrame(scores_filtered, index=problem['names'])
sns.set_style('white')
fig, ax = plt.subplots(1)
indices = Si_df[['S1', 'ST']]
err = Si_df[['S1_conf', 'ST_conf']]
indices.plot.bar(yerr=err.values.T, ax=ax)