def stack_mtslice(tdata_list):
"""stack tdata in the list into one instance"""
mean_temp_list = [tdata.mean_temp for tdata in tdata_list]
daily_temp_list = [tdata.daily_temp for tdata in tdata_list]
obs_mean_list = [tdata.obs_mean for tdata in tdata_list]
obs_std_list = [tdata.obs_std for tdata in tdata_list]
study_id_list = [tdata.study_id for tdata in tdata_list]
data_id_list = [tdata.data_id for tdata in tdata_list]
trimming_weights_list = [tdata.trimming_weights for tdata in tdata_list]
mean_temp = np.hstack(mean_temp_list)
daily_temp = np.hstack(daily_temp_list)
obs_mean = np.hstack(obs_mean_list)
obs_std = np.hstack(obs_std_list)
study_id = np.hstack(study_id_list)
data_id = np.hstack(data_id_list)
trimming_weights = np.hstack(trimming_weights_list)
if np.any(data_id == None):
data_id = None
return utils.TempData(mean_temp,
daily_temp,
obs_mean,
obs_std,
study_id,
data_id,
trimming_weights=trimming_weights)
def aggregate_mtslice(tdata):
"""aggregate data by the weighted mean"""
# extract the unique mean and daily pair
unique_pair = np.unique(np.vstack((tdata.mean_temp, tdata.daily_temp)).T,
axis=0)
mean_temp = unique_pair[:, 0]
daily_temp = unique_pair[:, 1]
obs_mean = []
obs_std = []
for p in unique_pair:
valid_id = (tdata.mean_temp == p[0]) &\
(tdata.daily_temp == p[1]) &\
(tdata.trimming_weights > 0.5)
obs_mean_atp = tdata.obs_mean[valid_id]
obs_std_atp = tdata.obs_std[valid_id]
ivar = 1.0 / obs_std_atp**2
obs_mean_atp = obs_mean_atp.dot(ivar) / np.sum(ivar)
obs_std_atp = np.sqrt(1.0 / np.sum(ivar))
obs_mean.append(obs_mean_atp)
obs_std.append(obs_std_atp)
obs_mean = np.array(obs_mean)
obs_std = np.array(obs_std)
study_id = np.arange(obs_mean.size)
data_id = None
return utils.TempData(mean_temp, daily_temp, obs_mean, obs_std, study_id,
data_id)
def extract_mtslice(tdata, mt):
"""extract the temperature data at given mean_temp"""
valid_id = tdata.mean_temp == mt
mean_temp = tdata.mean_temp[valid_id]
daily_temp = tdata.daily_temp[valid_id]
obs_mean = tdata.obs_mean[valid_id]
obs_std = tdata.obs_std[valid_id]
study_id = tdata.study_id[valid_id]
if tdata.data_id is not None:
data_id = tdata.data_id[valid_id]
else:
data_id = None
trimming_weights = tdata.trimming_weights[valid_id]
return utils.TempData(mean_temp,
daily_temp,
obs_mean,
obs_std,
study_id,
data_id,
trimming_weights=trimming_weights)
def load_data(path_to_data, outcome):
"""load data csv file"""
df = pd.read_csv(path_to_data)
mean_temp = df['meanTempDegree'].values
daily_temp = df['dailyTempCat'].values
obs_mean = df['lnRr_' + outcome].values
obs_std = df['se_' + outcome].values
study_id = df['adm1'].values
data_id = np.arange(df.shape[0])
valid_id = ~(np.isnan(obs_std) | np.isnan(obs_mean) | np.isinf(obs_std)
| np.isinf(obs_mean))
mean_temp = mean_temp[valid_id]
daily_temp = daily_temp[valid_id]
obs_mean = obs_mean[valid_id]
obs_std = obs_std[valid_id]
study_id = study_id[valid_id]
data_id = data_id[valid_id]
return utils.TempData(mean_temp, daily_temp, obs_mean, obs_std, study_id,
data_id)