def custom_interpolate(df):
"""Interpolate attributable burden draws"""
draw_cols = ['draw_{}'.format(i) for i in xrange(1000)]
id_cols = list(set(df.columns) - (set(draw_cols + ['year_id'])))
dfs = []
for year_range in [[1990, 1995], [1995, 2000], [2000, 2005], [2005, 2010]]:
start_df = (df.ix[df.year_id == year_range[0]].sort_values(id_cols)
.reset_index(drop=True))
end_df = (df.ix[df.year_id == year_range[1]].sort_values(id_cols)
.reset_index(drop=True))
ydf = interpolate(start_df, end_df, id_cols, 'year_id', draw_cols,
year_range[0], year_range[1])
dfs.append(ydf.query('year_id < {}'.format(year_range[1])))
df = pd.concat(dfs)
return df
def interp_loc(modelable_entity_id, measure_id, location_id, outpath):
start_year = 1980
epi_start_year = 1990
end_year = 2015
rank_year = 2005
# Retrieve epi draws and interpolate
epi_draws = []
for y in range(epi_start_year, end_year + 1, 5):
d = gopher.draws({'modelable_entity_ids': [modelable_entity_id]},
year_ids=[y],
location_ids=[location_id],
measure_ids=[measure_id],
verbose=False,
source="dismod",
age_group_ids=[
2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21
])
assert len(d) > 0, (
"Uh oh, couldn't find epi draws. Make sure you have "
"proportion estimates for the supplied meids")
epi_draws.append(d)
epi_draws = pd.concat(epi_draws)
ip_epi_draws = []
for y in range(epi_start_year, end_year, 5):
sy = y
ey = y + 5
ip_draws = maths.interpolate(
epi_draws.query('year_id==%s' % sy),
epi_draws.query('year_id==%s' % ey),
['age_group_id', 'model_version_id', 'sex_id'],
'year_id', ['draw_%s' % i for i in range(1000)],
sy,
ey,
rank_df=epi_draws.query('year_id==%s' % rank_year))
if ey != end_year:
ip_draws = ip_draws[ip_draws.year_id != ey]
ip_epi_draws.append(ip_draws)
ip_epi_draws = pd.concat(ip_epi_draws)
extrap_draws = []
for y in range(start_year, epi_start_year):
esy_draws = ip_epi_draws.query('year_id==%s' % epi_start_year)
esy_draws['year_id'] = y
extrap_draws.append(esy_draws)
epi_draws = pd.concat([ip_epi_draws] + extrap_draws)
epi_draws.to_csv(outpath)
def interpolate_ls(cvid, lid, sid):
modys = range(1990, 2016, 5)
iplys = [y for y in range(1990, 2016) if y not in modys]
idir = '/ihme/centralcomp/como/{cv}/draws/cause/total_interp/'.format(
cv=cvid)
try:
os.makedirs(idir)
except:
pass
moddfs = []
for y in modys:
moddfs.append(
pd.read_hdf('/ihme/centralcomp/como/{cv}/draws/cause/total/'
'3_{l}_{y}_{s}.h5'.format(cv=cvid, l=lid, y=y, s=sid)))
moddfs = pd.concat(moddfs)
for i in range(len(modys) - 1):
sy = modys[i]
ey = modys[i + 1]
print 'interpolating %s %s %s %s' % (lid, sid, sy, ey)
id_cols = [
'location_id', 'year_id', 'age_group_id', 'sex_id', 'cause_id'
]
time_col = 'year_id'
value_cols = ['draw_%s' % d for d in range(1000)]
x = maths.interpolate(moddfs.query('year_id == %s' % sy),
moddfs.query('year_id == %s' % ey),
id_cols,
time_col,
value_cols,
sy,
ey,
rank_df=moddfs.query('year_id == 2005'))
x = x[x.year_id.isin(iplys)]
for y in x.year_id.unique():
fn = '{id}/3_{l}_{y}_{s}.h5'.format(id=idir, l=lid, y=y, s=sid)
x.query('year_id == %s' % y).to_hdf(fn,
'draws',
mode='w',
format='table',
data_columns=id_cols)
def load_RSM(for_interp, interpd):
# compile risk-standardized mort
if not os.path.exists(for_interp):
dfs = []
for year_id in range(1990, 2011, 5) + [2016]:
print('Collecting ' + str(year_id))
df = data_formatter(year_id, locsdf)
dfs.append(df)
df = pd.concat(dfs)
df.to_hdf(for_interp,
key="data",
format="table",
data_columns=['location_id', 'year_id'])
else:
df = pd.read_hdf(for_interp)
# interpolate
draw_cols = ['draw_{}'.format(i) for i in xrange(1000)]
id_cols = ['location_id','age_group_id','sex_id','measure_id','metric_id','cause_id']
dfs = []
for year_range in [[1990, 1995], [1995, 2000], [2000, 2005], [2005, 2010], [2010, 2016]]:
print(year_range)
start_df = (df.ix[df.year_id == year_range[0]].sort_values(id_cols)
.reset_index(drop=True))
end_df = (df.ix[df.year_id == year_range[1]].sort_values(id_cols)
.reset_index(drop=True))
ydf = interpolate(start_df, end_df, id_cols, 'year_id', draw_cols,
year_range[0], year_range[1])
ydf = ydf.query('year_id < {} or year_id == 2016'.format(year_range[1]))
dfs.append(ydf)
df = pd.concat(dfs)
df.to_hdf(interpd,
key="data",
format="table",
data_columns=['location_id', 'year_id'])
return df