if "Phi" in trace.varnames:
self.Phi = trace.Phi
elif "Phi_1" in trace.varnames:
self.Phi = trace.Phi_1
if __name__ == "__main__":
folds = [['FR', 'GR', 'NL', 'BA', 'LV'], ['SE', 'DE', 'LT', 'MY', 'BG'],
['FI', 'DK', 'CZ', 'RS', 'BE'], ['NO', 'SK', 'IL', 'CH', 'ES'],
['ZA', 'MX', 'IT', 'IE', 'GE'], ['RO', 'PL', 'MA', 'HU', 'SI'],
['NZ', 'SG', 'PT', 'HR', 'EE']]
eval_fold = ['AL', 'AT', 'GB', 'AD', 'IS', 'MT']
dp = DataPreprocessor()
for fold_i, fold in enumerate(folds):
data = dp.preprocess_data(
"notebooks/double-entry-data/double_entry_final.csv",
last_day="2020-05-30",
schools_unis="whoops")
data.mask_reopenings()
r_is = []
for rg in fold:
c_s, d_s = mask_region(data, rg)
r_is.append(data.Rs.index(rg))
with cm_effect.models.CMCombined_Final_DifEffects(data, None) as model:
model.DailyGrowthNoise = args.growth_noise
model.RegionVariationNoise = args.country_noise
d_s = np.nonzero(np.cumsum(d.NewDeaths.data[i, :] > 0) == days + 1)[0]
if len(d_s) > 0:
d_s = d_s[0]
else:
d_s = len(d.Ds)
d.Active.mask[i, c_s:] = True
d.Confirmed.mask[i, c_s:] = True
d.Deaths.mask[i, d_s:] = True
d.NewDeaths.mask[i, d_s:] = True
d.NewCases.mask[i, c_s:] = True
if __name__ == "__main__":
dp = DataPreprocessor()
exp_num = args.exp
print(f"running exp {exp_num}")
# structural sensitivity
if exp_num == 1:
data = dp.preprocess_data(
"notebooks/double-entry-data/double_entry_final.csv",
last_day="2020-05-30",
schools_unis="whoops")
data.mask_reopenings()
with cm_effect.models.CMCombined_Additive(data, None) as model:
model.build_model()
elif exp_num == 2:
d.NewCases.mask[i, c_s:] = True
if __name__ == "__main__":
class ResultsObject():
def __init__(self, indxs, trace):
self.CMReduction = trace.CMReduction
self.RegionLogR = trace.RegionLogR[:, indxs]
self.InfectedCases = trace.InfectedCases[:, indxs, :]
self.InfectedDeaths = trace.InfectedDeaths[:, indxs, :]
self.ExpectedCases = trace.ExpectedCases[:, indxs, :]
self.ExpectedDeaths = trace.ExpectedDeaths[:, indxs, :]
self.Phi = trace.Phi_1
dp = DataPreprocessor(min_confirmed=100, drop_HS=True)
data = dp.preprocess_data("notebooks/final_data/data_final.csv")
HO_rs = ["DE", "PT", "CZ", "PL", "MX", "NL"]
indxs = [data.Rs.index(rg) for rg in HO_rs]
unmask_all(data)
for region in HO_rs:
mask_region(data, region)
print(f"Growth Noise {args.g}")
with cm_effect.models.CMCombined_Final_V3(data, None) as model:
model.DailyGrowthNoise = args.g
model.build_model()
with model.model:
model.trace = pm.sample(args.nS, chains=args.nC, target_accept=0.95)
if "Phi" in trace.varnames:
self.Phi = trace.Phi
elif "Phi_1" in trace.varnames:
self.Phi = trace.Phi_1
if __name__ == "__main__":
folds = [['DE', 'HU', 'FI', 'IE', 'RS', 'BE'],
['DK', 'GR', 'NO', 'FR', 'RO', 'MA'],
['ES', 'CZ', 'NL', 'CH', 'PT', 'AT'],
['IL', 'SE', 'IT', 'MX', 'GB', 'PL']]
fold_rs = folds[args.fold]
dp = DataPreprocessor(drop_HS=True)
data = dp.preprocess_data("notebooks/final_data/data_final.csv")
major_interventions = ["School Closure", "Stay Home Order", "Some Businesses Suspended",
"Most Businesses Suspended", "Gatherings <10", "Gatherings <1000", "Gatherings <100"]
minor_interventions = ["Mask Wearing", "Symptomatic Testing"]
ActiveCMs = copy.deepcopy(data.ActiveCMs)
maj_indxs = np.array([data.CMs.index(x) for x in major_interventions])
min_indxs = np.array([data.CMs.index(x) for x in minor_interventions])
nRs, nCMs, nDs = ActiveCMs.shape
for r in range(nRs):
maj_active = np.sum(data.ActiveCMs[r, maj_indxs, :], axis=0)
with cm_effect.models.CMCombined_ICL_NoNoise_3(data, None) as model:
model.build_model()
elif args.model == 2:
with cm_effect.models.CMCombined_ICL_NoNoise_4(data, None) as model:
model.build_model()
elif args.model == 3:
with cm_effect.models.CMCombined_ICL_NoNoise_5(data, None) as model:
model.build_model()
elif args.model == 4:
with cm_effect.models.CMCombined_ICL_NoNoise_6(data, None) as model:
model.build_model()
elif args.model == 5:
with cm_effect.models.CMCombined_ICL_NoNoise_7(data, None) as model:
model.build_model()
elif args.model == 6:
dp = DataPreprocessor(drop_HS=True)
dp.N_smooth = 1
data = dp.preprocess_data("notebooks/final_data/data_final.csv")
with cm_effect.models.CMCombined_ICL_NoNoise_7(data, None) as model:
model.build_model()
with model.model:
model.trace = pm.sample(1500, chains=6, target_accept=0.9)
np.savetxt(f"icl_bugs/model_{args.model+2}.csv", model.trace.CMReduction)
warnings.simplefilter(action="ignore", category=FutureWarning)
from epimodel.pymc3_models import cm_effect
from epimodel.pymc3_models.cm_effect.datapreprocessor import DataPreprocessor
import argparse
import pickle
argparser = argparse.ArgumentParser()
argparser.add_argument("--l", dest="last_date", type=int)
argparser.add_argument("--m", dest="model", type=int)
args = argparser.parse_args()
if __name__ == "__main__":
last_dates = ["2020-04-25", "2020-05-05", "2020-05-15", "2020-05-25", "2020-05-30"]
dp = DataPreprocessor()
data = dp.preprocess_data("notebooks/double-entry-data/double_entry_final.csv",
last_day=last_dates[args.last_date])
data.mask_reopenings()
if args.model == 0:
with cm_effect.models.CMCombined_Final(data, None) as model:
model.build_model(serial_interval_mean=6.7, serial_interval_sigma=2.1)
elif args.model == 1:
with cm_effect.models.CMCombined_Final(data, None) as model:
model.build_model(serial_interval_mean=5.1, serial_interval_sigma=1.8)
elif args.model == 2:
with cm_effect.models.CMCombined_Final(data, None) as model:
model.build_model(serial_interval_mean=6.68, serial_interval_sigma=4.88)
elif args.model == 3:
with cm_effect.models.CMCombined_Final_Reset1(data, None) as model: