def read_data_by_draw(artifact_path: str, key: str, draw: int) -> pd.DataFrame:
"""Reads data from the artifact on a per-draw basis. This
is necessary for Low Birthweight Short Gestation (LBWSG) data.
Parameters
----------
artifact_path
The artifact to read from.
key
The entity key associated with the data to read.
draw
The data to retrieve.
"""
key = key.replace(".", "/")
with pd.HDFStore(artifact_path, mode='r') as store:
index = store.get(f'{key}/index')
draw = store.get(f'{key}/draw_{draw}')
draw = draw.rename("value")
data = pd.concat([index, draw], axis=1)
data = data.drop(columns='location')
data = pivot_categorical(data)
data[project_globals.LBWSG_MISSING_CATEGORY.
CAT] = project_globals.LBWSG_MISSING_CATEGORY.EXPOSURE
return data
def _get_excess_shift_source(self, builder: Builder) -> LookupTable:
excess_shift_data = builder.data.load(
f'{self.risk}.excess_shift',
affected_entity=self.target.name,
affected_measure=self.target.measure)
excess_shift_data = rebin_relative_risk_data(builder, self.risk,
excess_shift_data)
excess_shift_data = pivot_categorical(excess_shift_data)
return builder.lookup.build_table(excess_shift_data,
key_columns=['sex'],
parameter_columns=['age', 'year'])
def get_relative_risk_data(self, builder):
relative_risk_data = read_data_by_draw(builder,
f'{self.risk}.relative_risk')
correct_target = ((relative_risk_data['affected_entity'] == 'all')
& (relative_risk_data['affected_measure']
== 'excess_mortality_rate'))
relative_risk_data = (relative_risk_data[correct_target].drop(
['affected_entity', 'affected_measure'], 'columns'))
relative_risk_data = pivot_categorical(relative_risk_data)
relative_risk_data[project_globals.LBWSG_MISSING_CATEGORY.CAT] = (
relative_risk_data['cat106'] + relative_risk_data['cat116']) / 2
return relative_risk_data
def setup(self, builder: 'Builder'):
self.clock = builder.time.clock()
self.randomness = builder.randomness.get_stream(f'{self.name}_initial_states')
disability_weight_data = builder.data.load(project_globals.VITAMIN_A_DEFICIENCY_DISABILITY_WEIGHT)
self.base_disability_weight = builder.lookup.build_table(disability_weight_data,
key_columns=['sex'],
parameter_columns=['age', 'year'])
self.disability_weight = builder.value.register_value_producer(
f'{self.name}.disability_weight',
source=self.compute_disability_weight,
requires_columns=['age', 'sex', 'alive', self.name])
builder.value.register_value_modifier('disability_weight', modifier=self.disability_weight)
exposure_data = builder.data.load(project_globals.VITAMIN_A_DEFICIENCY_EXPOSURE)
exposure_data = pivot_categorical(exposure_data)
exposure_data = exposure_data.drop('cat2', axis=1)
self._base_exposure = builder.lookup.build_table(exposure_data,
key_columns=['sex'],
parameter_columns=['age', 'year'])
self.exposure_proportion = builder.value.register_value_producer(
f'{self.name}.exposure_parameters',
source=self.exposure_proportion_source,
requires_values=[f'{self.name}.exposure_parameters.paf'],
requires_columns=['age', 'sex']
)
base_paf = builder.lookup.build_table(0)
self.joint_paf = builder.value.register_value_producer(
f'{self.name}.exposure_parameters.paf',
source=lambda index: [base_paf(index)],
preferred_combiner=list_combiner,
preferred_post_processor=union_post_processor
)
self.exposure = builder.value.register_value_producer(
f'{self.name}.exposure',
source=self.get_current_exposure,
requires_columns=['age', 'sex', f'{self.name}_propensity']
)
columns_created = [self.name,
project_globals.VITAMIN_A_GOOD_EVENT_TIME, project_globals.VITAMIN_A_GOOD_EVENT_COUNT,
project_globals.VITAMIN_A_BAD_EVENT_TIME, project_globals.VITAMIN_A_BAD_EVENT_COUNT,
project_globals.VITAMIN_A_PROPENSITY]
view_columns = columns_created + ['alive', 'age', 'sex']
self.population_view = builder.population.get_view(view_columns)
builder.population.initializes_simulants(
self.on_initialize_simulants,
creates_columns=columns_created,
requires_columns=['age', 'sex'],
requires_streams=[f'{self.name}_initial_states']
)
builder.event.register_listener('time_step', self.on_time_step)
def prep_weights(art):
weights = art.load(
'risk_factor.high_systolic_blood_pressure.exposure_distribution_weights'
)
index_cols = weights.index.names.difference({'location', 'parameter'})
weights = pivot_categorical(weights.reset_index())
weights = weights.drop(columns=index_cols)
if 'glnorm' in weights.columns:
if np.any(weights['glnorm']):
raise NotImplementedError('glnorm distribution is not supported')
weights = weights.drop(columns='glnorm')
return weights
def load_relative_risk_data(self, builder):
relative_risk_data = builder.data.load(f'{self.risk}.relative_risk')
correct_target = ((relative_risk_data['affected_entity']
== project_globals.RATE_TARGET_MAP[self.target].name)
& (relative_risk_data['affected_measure']
== project_globals.RATE_TARGET_MAP[self.target].measure))
relative_risk_data = (relative_risk_data[correct_target]
.drop(['affected_entity', 'affected_measure'], 'columns'))
if get_distribution_type(builder, self.risk) in ['dichotomous', 'ordered_polytomous', 'unordered_polytomous']:
relative_risk_data = pivot_categorical(relative_risk_data)
else:
relative_risk_data = relative_risk_data.drop(['parameter'], 'columns')
return relative_risk_data
def get_relative_risk_data_by_draw(builder, risk: EntityString,
target: TargetString,
randomness: RandomnessStream):
source_type = validate_relative_risk_data_source(builder, risk, target)
relative_risk_data = load_relative_risk_data_by_draw(
builder, risk, target, source_type)
relative_risk_data = rebin_relative_risk_data(builder, risk,
relative_risk_data)
if get_distribution_type(builder, risk) in [
'dichotomous', 'ordered_polytomous', 'unordered_polytomous'
]:
relative_risk_data = pivot_categorical(relative_risk_data)
else:
relative_risk_data = relative_risk_data.drop(['parameter'], 'columns')
return relative_risk_data
def get_exposure_data(builder):
exposure = read_data_by_draw(builder, project_globals.LBWSG_EXPOSURE)
exposure = pivot_categorical(exposure)
exposure[project_globals.LBWSG_MISSING_CATEGORY.
CAT] = project_globals.LBWSG_MISSING_CATEGORY.EXPOSURE
return exposure