def _load_model_for_region(self, scenario="inferred"):
"""
Try to load a model for the region, else load the state level model and update parameters
for the region.
"""
model = self.regional_input.load_mle_fit_model()
if model:
inferred_params = self.regional_input.load_inference_result()
else:
_log.info(f"No MLE model found. Reverting to state level.",
region=self.regional_input.region)
model = self.regional_input.load_state_mle_fit_model()
if model:
inferred_params = self.regional_input.load_state_inference_result(
)
else:
raise FileNotFoundError(
f"Could not locate state result for {self.state_name}")
# Rescale state values to the county population and replace county
# specific params.
# TODO: get_average_seir_parameters should return the analytic solution when available
# right now it runs an average over the ensemble (with N_samples not consistently set
# across the code base).
default_params = ParameterEnsembleGenerator(
N_samples=500,
combined_datasets_latest=self.regional_input.latest,
t_list=model.t_list,
suppression_policy=model.suppression_policy,
).get_average_seir_parameters()
population_ratio = default_params["N"] / model.N
model.N *= population_ratio
model.I_initial *= population_ratio
model.E_initial *= population_ratio
model.A_initial *= population_ratio
model.S_initial = model.N - model.I_initial - model.E_initial - model.A_initial
for key in {"beds_general", "beds_ICU", "ventilators"}:
setattr(model, key, default_params[key])
# Determine the appropriate future suppression policy based on the
# scenario of interest.
eps_final = sp.estimate_future_suppression_from_fits(inferred_params,
scenario=scenario)
model.suppression_policy = sp.get_epsilon_interpolator(
eps=inferred_params["eps"],
t_break=inferred_params["t_break"],
eps2=inferred_params["eps2"],
t_delta_phases=inferred_params["t_delta_phases"],
t_break_final=(datetime.datetime.today() -
datetime.datetime.fromisoformat(
inferred_params["t0_date"])).days,
eps_final=eps_final,
)
model.run()
return model
def run_model(self, R0, eps, t_break, eps2, t_delta_phases,
log10_I_initial):
"""
Generate the model and run.
Parameters
----------
R0: float
Basic reproduction number
eps: float
Fraction of reduction in contact rates in the second stage.
t_break: float
Timing for the switch in suppression policy.
eps2: float
Fraction of reduction in contact rates in the third stage
t_delta_phases: float
Timing for the switch in from second to third stage.
log10_I_initial:
log10 initial infections.
Returns
-------
model: SEIRModel
The SEIR model that has been run.
"""
suppression_policy = suppression_policies.get_epsilon_interpolator(
eps, t_break, eps2, t_delta_phases)
if self.with_age_structure:
age_distribution = self.SEIR_kwargs["N"] / self.SEIR_kwargs[
"N"].sum()
seir_model = SEIRModelAge
else:
age_distribution = 1
seir_model = SEIRModel
# Load up some number of initial exposed so the initial flow into
# infected is stable.
self.SEIR_kwargs["E_initial"] = (
self.steady_state_exposed_to_infected_ratio * 10**log10_I_initial *
age_distribution)
model = seir_model(
R0=R0,
suppression_policy=suppression_policy,
I_initial=10**log10_I_initial * age_distribution,
**self.SEIR_kwargs,
)
model.run()
return model
def _load_model_for_fips(self, scenario="inferred"):
"""
Try to load a model for the locale, else load the state level model
and update parameters for the county.
"""
artifact_path = get_run_artifact_path(self.fips, RunArtifact.MLE_FIT_MODEL)
if os.path.exists(artifact_path):
with open(artifact_path, "rb") as f:
model = pickle.load(f)
inferred_params = fit_results.load_inference_result(self.fips)
else:
_logger.info(
f"No MLE model found for {self.state_name}: {self.fips}. Reverting to state level."
)
artifact_path = get_run_artifact_path(self.fips[:2], RunArtifact.MLE_FIT_MODEL)
if os.path.exists(artifact_path):
with open(artifact_path, "rb") as f:
model = pickle.load(f)
inferred_params = fit_results.load_inference_result(self.fips[:2])
else:
raise FileNotFoundError(f"Could not locate state result for {self.state_name}")
# Rescale state values to the county population and replace county
# specific params.
# TODO: get_average_seir_parameters should return the analytic solution when available
# right now it runs an average over the ensemble (with N_samples not consistently set
# across the code base).
default_params = ParameterEnsembleGenerator(
self.fips,
N_samples=500,
t_list=model.t_list,
suppression_policy=model.suppression_policy,
).get_average_seir_parameters()
population_ratio = default_params["N"] / model.N
model.N *= population_ratio
model.I_initial *= population_ratio
model.E_initial *= population_ratio
model.A_initial *= population_ratio
model.S_initial = model.N - model.I_initial - model.E_initial - model.A_initial
for key in {"beds_general", "beds_ICU", "ventilators"}:
setattr(model, key, default_params[key])
# Determine the appropriate future suppression policy based on the
# scenario of interest.
eps_final = sp.estimate_future_suppression_from_fits(inferred_params, scenario=scenario)
model.suppression_policy = sp.get_epsilon_interpolator(
eps=inferred_params["eps"],
t_break=inferred_params["t_break"],
eps2=inferred_params["eps2"],
t_delta_phases=inferred_params["t_delta_phases"],
t_break_final=(
datetime.datetime.today()
- datetime.datetime.fromisoformat(inferred_params["t0_date"])
).days,
eps_final=eps_final,
)
model.run()
return model