def summarize_loc(source,
drawdir,
outdir,
location_id,
year_id,
rei_ids,
change_intervals=None,
gbd_round_id=5):
'''summarize every rei for a single location'''
# Set global age weights
gbd_round_map = get_ids('gbd_round')
gbd_round = gbd_round_map.loc[gbd_round_map.gbd_round_id ==
gbd_round_id].gbd_round.item()
Globals.aw = get_age_weights(gbd_round_id=int(gbd_round_id))
# Set global population
pops = []
popfiles = glob(os.path.join(drawdir, 'population_*.csv'))
for popfile in popfiles:
pops.append(pd.read_csv(popfile))
pops = pd.concat(pops).drop_duplicates(
subset=['location_id', 'age_group_id', 'year_id', 'sex_id'])
Globals.pop = pops.rename(columns={'population': 'pop_scaled'})
pool = Pool(10)
results = pool.map(summ_loc, [((source, location_id, rei, year_id,
change_intervals, gbd_round_id), {})
for rei in rei_ids])
pool.close()
pool.join()
results = [res for res in results if isinstance(res, tuple)]
results = list(zip(*results))
single_year = pd.concat([res for res in results[0] if res is not None])
single_year = single_year[[
'rei_id', 'location_id', 'year_id', 'age_group_id', 'sex_id',
'measure_id', 'metric_id', 'val', 'lower', 'upper'
]]
single_file = os.path.join(outdir,
'single_year_{}.csv'.format(location_id))
single_year.to_csv(single_file, index=False)
os.chmod(single_file, 0o775)
multi_year = pd.concat(results[1])
if len(multi_year) > 0:
multi_year = multi_year[[
'rei_id', 'location_id', 'year_start_id', 'year_end_id',
'age_group_id', 'sex_id', 'measure_id', 'metric_id', 'val',
'lower', 'upper'
]]
multi_year.replace([np.inf, -np.inf], np.nan)
multi_year.dropna(inplace=True)
multi_file = os.path.join(outdir,
'multi_year_{}.csv'.format(location_id))
multi_year.to_csv(multi_file, index=False)
os.chmod(multi_file, 0o775)
def _cache_age_weights(self) -> None:
logger.debug(
"Starting to load age_weights cache")
age_weights_df = get_age_weights(gbd_round_id=int(self.gbd_round_id))
# returns age_group_id, age_group_weight_value as a pandas df
cache_file = "FILEPATH".format(
self.cache_dir)
age_weights_df.to_hdf(cache_file, "age_weights",
data_columns=['age_group_id'], format="table")
logger.debug(
"Cached age_weights in {}".format(cache_file))
def summarize_loc(source,
drawdir,
outdir,
location_id,
year_id,
rei_ids,
change_intervals,
gbd_round_id):
'''summarize every rei for a single location'''
# Set global age weights
aw = get_age_weights(gbd_round_id=int(gbd_round_id))
# Set global population
pops = pd.read_hdf("FILEPATH/population.h5")
pops = pops.loc[pops.location_id == location_id]
pop = pops.rename(columns={'population': 'pop_scaled'})
pool = Pool(10)
results = pool.map(summ_loc, [(
(source, location_id, rei, year_id, change_intervals, gbd_round_id,
pop, aw), {})
for rei in rei_ids])
pool.close()
pool.join()
results = [res for res in results if isinstance(res, tuple)]
results = list(zip(*results))
single_year = pd.concat([res for res in results[0] if res is not None],sort=True)
single_year = single_year[
['rei_id', 'location_id', 'year_id', 'age_group_id', 'sex_id',
'measure_id', 'metric_id', 'val', 'lower', 'upper']]
single_file = os.path.join(outdir, 'single_year_{}.csv'.format(location_id))
single_year.to_csv(single_file, index=False)
os.chmod(single_file, 0o775)
multi_year = pd.concat(results[1],sort=True)
if len(multi_year) > 0:
multi_year = multi_year[
['rei_id', 'location_id', 'year_start_id', 'year_end_id',
'age_group_id', 'sex_id', 'measure_id', 'metric_id', 'val',
'lower', 'upper']]
multi_year.replace([np.inf, -np.inf], np.nan)
multi_year.dropna(inplace=True)
multi_file = os.path.join(outdir, 'multi_year_{}.csv'.format(location_id))
multi_year.to_csv(multi_file, index=False)
os.chmod(multi_file, 0o775)
def _compute_age_standardized_rate(
data: pd.DataFrame,
gbd_round_id: int
) -> pd.DataFrame:
"""
Computes the age standardized rate from most-detailed age data and returns
as a new dataframe.
Arguments:
data (pd.DataFrame)
gbd_round_id (int)
Returns:
pd.DataFrame
"""
age_std: pd.DataFrame = data.copy()
age_weights: pd.DataFrame = get_age_weights(gbd_round_id=gbd_round_id)
# Merge age-weight data
age_std = pd.merge(
age_std,
age_weights,
on=[Columns.AGE_GROUP_ID],
how='left'
)
# calculate age-standardized value:
# first we divide all draws by the population value for the respective
# demographic, then we multiply the result by the age weight.
age_std[Columns.DRAWS] = age_std[Columns.DRAWS].div(
age_std[Columns.POPULATION].values, axis='index'
).mul(age_std[Columns.AGE_WEIGHT_VALUE].values, axis='index')
# set all age groups to age-standardized
age_std[Columns.AGE_GROUP_ID] = gbd.age.AGE_STANDARDIZED
age_std.drop(Columns.AGE_WEIGHT_VALUE, inplace=True, axis=1)
# groupby sum to aggregate into final age-standardized rates
age_std = age_std.groupby(Columns.INDEX).sum().reset_index()
# mark the metric for these estimates as rate
age_std[Columns.METRIC_ID] = gbd.metrics.RATE
return age_std
def std_age_weights(self):
if self._std_age_weights is None:
self._std_age_weights = get_age_weights(
gbd_round_id=self.gbd_round_id)
return self._std_age_weights