## Encode data
from load_diabetes import fetch_data
male_df, male_df, data_dtypes = fetch_data()
data_dtypes['G03.DDD'] = 'int64'
male_N = len(male_df)
male_N = len(male_df)
##################################################
### Define model ###
## For male
# Load variable type dictionaries for both independent and dependent types
from variable_types import independent_model as male_variable_types_
male_variable_types_.pop('G03.DDD')
dead_male_variable_types = male_variable_types_.copy()
dead_male_variable_types.pop('dead')
alive_male_variable_types = dead_male_variable_types.copy()
alive_male_variable_types.pop('ep')
alive_male_variable_types.pop('lex.dur')
# Pick features for training
male_features = list(male_variable_types_.keys())
male_features.remove('pi_unconstrained')
# Cast features to appropriate dtypes
male_dtypes = {key:value if value!='O' else 'int64' for key, value in \
data_dtypes[male_features].items()}
alive_features = list(alive_male_variable_types.keys())
alive_features.remove('pi_unconstrained')
female_df, male_df, data_dtypes = fetch_data()
data_dtypes['G03.DDD'] = 'int64'
data_dtypes['is.female'] = 'int64'
female_df["is.female"] = 1
male_df["is.female"] = 0
train_df = pd.concat([female_df, male_df])
N = len(train_df)
##################################################
### Define model ###
## For male
# Load variable type dictionaries for both independent and dependent types
from variable_types import independent_model as train_variable_types_
train_variable_types = train_variable_types_.copy()
train_variable_types.pop('dead')
train_variable_types["is.female"] = "Bernoulli"
# Pick features for training
train_features = list(train_variable_types.keys())
train_features.remove('pi_unconstrained')
# Cast features to appropriate dtypes
train_dtypes = {key:value if value!='O' else 'int64' for key, value in \
data_dtypes[train_features].items()}
# Pick features
train_df = train_df[train_features]
def main():
### Load diabetes data ###
## Encode data
from load_diabetes import fetch_data
female_df, male_df, data_dtypes = fetch_data()
data_dtypes['G03.DDD'] = 'int64'
female_N = len(female_df)
male_N = len(male_df)
##################################################
### Define model ###
## For female
# Load variable type dictionaries for both independent and dependent types
from variable_types import independent_model as female_variable_types_
dead_female_variable_types = female_variable_types_.copy()
dead_female_variable_types.pop('dead')
alive_female_variable_types = dead_female_variable_types.copy()
alive_female_variable_types.pop('ep')
alive_female_variable_types.pop('lex.dur')
# Pick features for training
female_features = list(female_variable_types_.keys())
female_features.remove('pi_unconstrained')
# Cast features to appropriate dtypes
female_dtypes = {key:value if value!='O' else 'int64' for key, value in \
data_dtypes[female_features].items()}
alive_features = list(alive_female_variable_types.keys())
alive_features.remove('pi_unconstrained')
from collections import OrderedDict
npr.seed(1234)
use_cuda = False
if use_cuda:
torch.set_default_tensor_type('torch.cuda.DoubleTensor')
torch.cuda.manual_seed(1234)
else:
torch.set_default_tensor_type('torch.DoubleTensor')
torch.manual_seed(1234)
##################################################
##################################################
from variable_types import independent_model as train_variable_types_
train_variable_types_base = train_variable_types_.copy()
train_variable_types_base.pop('dead')
train_variable_types_base["is.female"] = "Bernoulli"
maps = pickle.load(open('maps.pickle', 'rb'))[0]
maps['age'] = lambda x: sum(maps['age_lim'] * np.array([-1, 1])) * x + maps[
'age_lim'][0]
maps['per'] = lambda x: sum(maps['per_lim'] * np.array([-1, 1])) * x + maps[
'per_lim'][0]
from sampler import fast_sample
from load_diabetes import decode_data
N_female = 208148
N_male = 226372
N = N_female + N_male