def test_Ensemble_data():
"""tests to check data from an ensemble"""
# Test that the ensemble of two identical models
# behalves like twice a single model.
# Only for the independent (ie. diagonal) case
# Note: this test would fail if a fixed contact_matrix
# is passed which happens to be the identity matrix
# The difference is that when specified as independent
# each model is booted independently
# If a contact matrix is specified, then the boot goal
# is the combined total of all models. So there will be a different starting point.
test_a = Model.open_file(path_model_2)
test_a.name = 'test_a'
test_b = Model.open_file(path_model_2)
test_b.name = 'test_b'
test_b.parameters['alpha_0'].set_value(0.7)
reference = Model.open_file(path_model_2)
reference.name = 'reference'
off_diag = Parameter('off_diagonal',
0.1,
parameter_min=0.,
parameter_max=1.,
description='off diagonal element of contact matrix')
off_diags = [off_diag]
test_ensemble = Ensemble('test_ensemble', reference)
test_ensemble.upload_models([test_a, test_b])
test_ensemble.define_cross_transmission('infection cycle',
'infected',
'susceptible',
'total',
'contagious',
'alpha',
contact_type='simple',
contact=off_diags)
n_days = 100
norm_day = 50
test_ensemble.reset()
test_ensemble.evolve_expectations(norm_day)
for key in test_ensemble.populations:
pop = test_ensemble.populations[key]
nu = pop.history[norm_day]
pop.history[norm_day] = int(round(nu))
pop.scale_future(1., expectations=False)
for model_name in test_ensemble.models:
model = test_ensemble.models[model_name]
for pop_name in model.populations:
pop = model.populations[pop_name]
nu = pop.history[norm_day]
pop.history[norm_day] = int(round(nu))
pop.scale_future(1., expectations=False)
test_ensemble.generate_data(n_days, norm_day)
for pop_name in test_ensemble.populations:
pop = test_ensemble.populations[pop_name]
if pop.show_sim:
ens_hist = test_ensemble.populations[pop_name].history
def test_point_estimate_local_death():
start_day = 80
end_day = 100
ref_2 = Model.open_file(
'/Users/karlen/pypm-local/models/covid19/USA/ny_4_2_0109.pypm')
sim_2 = Model.open_file(
'/Users/karlen/pypm-local/models/covid19/USA/ny_4_2_0109.pypm')
# do fit of recover_frac
for par_name in ref_2.parameters:
par = ref_2.parameters[par_name]
par.set_fixed()
for par_name in ['recover_frac']:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
sim_2.reset()
sim_2.generate_data(end_day)
optimizer = Optimizer(ref_2,
'total deaths',
sim_2.populations['deaths'].history,
[start_day, end_day],
cumul_reset=True)
optimizer.reset_variables()
popt, pcov = optimizer.fit()
iii = 1
def test_point_estimates_repeated():
start_day = 12
end_day = 60
ref_2 = Model.open_file(path_model_2_2)
sim_2 = Model.open_file(path_model_2_2)
# do fit of alpha_0, alpha_1, cont_0
par_names = ['alpha_0', 'alpha_1', 'cont_0']
sums = {}
sum2s = {}
for par_name in par_names:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
sums[par_name] = 0.
sum2s[par_name] = 0.
n_rep = 10
fit_stat_list = []
for i in range(n_rep):
sim_2.reset()
sim_2.generate_data(end_day)
optimizer = Optimizer(ref_2, 'total reported',
sim_2.populations['reported'].history,
[start_day, end_day])
optimizer.reset_variables()
popt, pcov = optimizer.fit()
fit_stat_list.append(optimizer.fit_statistics)
for par_name in par_names:
value = ref_2.parameters[par_name].get_value()
sums[par_name] += value
sum2s[par_name] += value**2
ass_std = {}
ass_std['alpha_0'] = 0.03
ass_std['alpha_1'] = 0.01
ass_std['cont_0'] = 10.
means = {}
std = {}
for par_name in par_names:
means[par_name] = sums[par_name] / n_rep
std[par_name] = np.sqrt(sum2s[par_name] / n_rep - means[par_name]**2)
assert std[par_name] < ass_std[par_name]
truth = ref_2.parameters[par_name].initial_value
assert np.abs((means[par_name] - truth) / std[par_name] /
np.sqrt(1. * n_rep)) < 3.
ndof = fit_stat_list[0]['ndof']
chi2_list = [fit_stat_list[i]['chi2'] for i in range(n_rep)]
chi2_mean = np.mean(chi2_list)
assert np.abs(chi2_mean - ndof) < 8.
acor_list = [fit_stat_list[i]['acor'] for i in range(n_rep)]
acor_mean = np.mean(acor_list)
assert np.abs(acor_mean) < 0.2
def test_Ensemble_properties_different():
"""tests to ensure the properties of Ensemble with different sub models"""
# Test that the ensemble of two identical models
# behalves like twice a single model.
# Only for the independent (ie. diagonal) case
# Note: this test would fail if a fixed contact_matrix
# is passed which happens to be the identity matrix
# The difference is that when specified as independent
# each model is booted independently
# If a contact matrix is specified, then the boot goal
# is the combined total of all models. So there will be a different starting point.
test_a = Model.open_file(path_model_2)
test_a.name = 'test_a'
test_b = Model.open_file(path_model_2)
test_b.name = 'test_b'
test_b.parameters['alpha_0'].set_value(0.7)
test_c = Model.open_file(path_model_2)
test_c.name = 'test_c'
test_d = Model.open_file(path_model_2)
test_d.name = 'test_d'
test_d.parameters['alpha_0'].set_value(0.7)
reference = Model.open_file(path_model_2)
reference.name = 'reference'
test_ensemble = Ensemble('test_ensemble', reference)
test_ensemble.upload_models([test_a, test_b])
test_ensemble.define_cross_transmission('infection cycle',
'infected',
'susceptible',
'total',
'contagious',
'alpha',
contact_type='diagonal')
n_days = 100
test_c.reset()
test_c.evolve_expectations(n_days)
test_d.reset()
test_d.evolve_expectations(n_days)
test_ensemble.reset()
test_ensemble.evolve_expectations(n_days)
for pop_name in test_ensemble.populations:
pop = test_ensemble.populations[pop_name]
if pop.show_sim:
ens_hist = test_ensemble.populations[pop_name].history
tc_hist = test_c.populations[pop_name].history
td_hist = test_d.populations[pop_name].history
for i in range(len(ens_hist)):
ratio = ens_hist[i] / (tc_hist[i] + td_hist[i])
assert np.abs(ratio - 1.) < 0.01
def test_Ensemble_properties_identical():
"""tests to ensure the properties of Ensemble"""
# Test that the ensemble of two identical models
# behalves like twice a single model.
# independent of the contact matrix
test_a = Model.open_file(path_model_2)
test_a.name = 'test_a'
test_b = Model.open_file(path_model_2)
test_b.name = 'test_b'
reference = Model.open_file(path_model_2)
reference.name = 'reference'
single = Model.open_file(path_model_2)
single.name = 'single'
test_ensemble = Ensemble('test_ensemble', reference)
test_ensemble.upload_models([test_a, test_b])
contacts = [
[[1., 0.0], [0.0, 1.]],
[[1., 1.0], [1.0, 1.]],
[[1., 0.5], [0.7, 1.]],
]
for contact in contacts:
test_ensemble.define_cross_transmission('infection cycle',
'infected',
'susceptible',
'total',
'contagious',
'alpha',
contact_type='fixed',
contact=contact)
n_days = 100
test_ensemble.reset()
test_ensemble.evolve_expectations(n_days)
single.reset()
single.evolve_expectations(n_days)
for pop_name in test_ensemble.populations:
pop = test_ensemble.populations[pop_name]
if pop.show_sim:
ens_hist = test_ensemble.populations[pop_name].history
single_hist = single.populations[pop_name].history
for i in range(len(ens_hist)):
ratio = ens_hist[i] / single_hist[i]
assert np.abs(ratio - 2.) < 0.001
def test_model_2_9():
ref_2_9 = Model.open_file(path_model_2_9)
ref_2_9.parameters['alpha_0'].set_value(0.35)
ref_2_9.transitions['outbreak_v'].enabled = True
ref_2_9.parameters['outbreak_v_time'].set_value(30)
ref_2_9.parameters['outbreak_v_number'].set_value(2.)
ref_2_9.parameters['alpha_0_v'].set_value(0.35)
ref_2_9.transitions['outbreak_w'].enabled = True
ref_2_9.parameters['outbreak_w_time'].set_value(60)
ref_2_9.parameters['outbreak_w_number'].set_value(4.)
ref_2_9.parameters['alpha_0_w'].set_value(0.45)
ref_2_9.transitions['vaccination_1'].enabled = True
ref_2_9.parameters['vacc_time_1'].set_value(80)
ref_2_9.parameters['vacc_number_1'].set_value(1000.)
# ref_2_9.reset()
# ref_2_9.evolve_expectations(200)
ref_2_9.reset()
ref_2_9.generate_data(200)
i = 1
def test_interval_maker():
hub_date = datetime.date(2020, 4, 1)
my_IntervalMaker = IntervalMaker("USA", hub_date)
categories = ['case', 'death', 'hospitalization']
n_period_dict = {'case': 5, 'death': 5, 'hospitalization': 30}
n_rep = 10
scale_std_alpha = 2.
model = Model.open_file(path_model_2_6)
if 'interval_maker' not in model.user_dict:
model.user_dict['interval_maker'] = {}
model.user_dict['interval_maker']['smearing parameters'] = [
'non_icu_hosp_frac', 'recover_frac'
]
model.parameters['non_icu_hosp_frac'].std_estimator = 0.002
model.parameters['recover_frac'].set_value(0.99)
model.parameters['recover_frac'].std_estimator = 0.01
my_IntervalMaker.get_quantiles(categories,
n_period_dict,
model,
n_rep=n_rep,
scale_std_alpha=scale_std_alpha,
back_up=21,
fall_back=True,
rescale=True)
for category in categories:
my_IntervalMaker.append_user_dict(category, model)
i = 1
def test_point_estimate_daily():
def delta(cumul):
diff = []
for i in range(1, len(cumul)):
diff.append(cumul[i] - cumul[i - 1])
# first daily value is repeated since val(t0-1) is unknown
diff.insert(0, diff[0])
return diff
start_day = 12
end_day = 60
ref_2 = Model.open_file(path_model_2_2)
sim_2 = Model.open_file(path_model_2_2)
# do fit of alpha_0, alpha_1, cont_0, trans_rate_1_time
for par_name in ['alpha_0', 'alpha_1', 'cont_0']:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
par = ref_2.parameters['trans_rate_1_time']
par.set_variable(None, None)
par.set_min(13)
par.set_max(19)
sim_2.reset()
sim_2.generate_data(end_day)
daily_data = delta(sim_2.populations['reported'].history)
daily_data[47] = np.inf
optimizer = Optimizer(ref_2,
'daily reported',
daily_data, [start_day, end_day],
skip_data='42,45:48')
optimizer.reset_variables()
scan_dict = optimizer.i_fit()
assert ref_2.parameters['trans_rate_1_time'].get_value() in [15, 16, 17]
par = ref_2.parameters['trans_rate_1_time']
par.set_fixed()
popt, pcov = optimizer.fit()
assert np.abs(ref_2.parameters['alpha_0'].get_value() -
ref_2.parameters['alpha_0'].initial_value) < 0.06
assert np.abs(ref_2.parameters['alpha_1'].get_value() -
ref_2.parameters['alpha_1'].initial_value) < 0.02
assert np.abs(ref_2.parameters['cont_0'].get_value() -
ref_2.parameters['cont_0'].initial_value) < 20.
def test_Model_properties():
"""tests to ensure the properties of Model"""
ref_model_1 = Model.open_file(path_model_4_1)
ref_model_2 = Model.open_file(path_model_4_1)
for test_model in [ref_model_1, ref_model_2]:
# check simple scaling of initial contagious population
# Will not be exact due to bootstrap
EPS = 0.01
cont_0 = test_model.parameters['cont_0'].get_value()
test_model.reset()
test_model.evolve_expectations(200)
n_reported1 = test_model.populations['reported'].history[-1]
assert test_model.populations['contagious_o'].history[-1] < 10.
test_model.parameters['cont_0'].set_value(2 * cont_0)
test_model.reset()
test_model.evolve_expectations(200)
n_reported2 = test_model.populations['reported'].history[-1]
assert np.abs(n_reported2 - n_reported1 * 2) < EPS * n_reported1 * 2
# check that the mean of many data runs is near expectation
ref_model = Model.open_file(path_model_2)
ref_model.reset()
n_days = 60
ref_model.evolve_expectations(n_days)
sim_models = []
n_rep = 100
for i in range(n_rep):
sim_model = copy.deepcopy(ref_model)
sim_model.reset()
sim_model.generate_data(n_days)
sim_models.append(sim_model)
for pop_name in ref_model.populations:
pop = ref_model.populations[pop_name]
if pop.show_sim:
results = []
for sim_model in sim_models:
results.append(sim_model.populations[pop_name].history[-1])
mean = np.mean(np.array(results))
std = np.std(np.array(results))
error = std / np.sqrt(1. * n_rep)
expect = ref_model.populations[pop_name].history[-1]
assert np.abs(expect - mean) < 8. * error
def test_max_mult():
ref_2_5 = Model.open_file(path_model_2_5)
ref_2_5.transitions['vaccination'].enabled = True
ref_2_5.parameters['vaccination_number'].set_value(800000.)
ref_2_5.parameters['vaccination_time'].set_value(5)
ref_2_5.reset()
ref_2_5.boot(expectations=True)
ref_2_5.evolve_expectations(200)
i = 1
def test_trajectory():
ref_2 = Model.open_file(path_model_2_2)
trajectory = Trajectory(ref_2, 'contagious', 'trans_rate_1', [0.03, 0.75])
alpha_c = trajectory.get_alpha(0.)
delta_1 = trajectory.get_delta(0.1)
delta_2 = trajectory.get_delta(0.2)
assert np.abs(alpha_c - 0.152) < 0.001
assert np.abs(trajectory.get_delta(alpha_c)) < 0.00001
assert np.abs(delta_1 + 0.0513) < 0.0001
assert np.abs(delta_2 - 0.0389) < 0.0001
def test_model_2_5():
ref_2_5 = Model.open_file(path_model_2_5)
ref_2_3 = Model.open_file(path_model_2_3)
#ref_2_5.transitions['vaccination'].enabled = True
#ref_2_5.parameters['vaccination_number'].set_value(10000.)
ref_2_5.reset()
ref_2_5.evolve_expectations(200)
ref_2_3.reset()
ref_2_3.evolve_expectations(200)
for pop_name in ref_2_3.populations:
pop = ref_2_3.populations[pop_name]
if pop.show_sim:
print(pop_name)
hist_2_3 = ref_2_3.populations[pop_name].history
hist_2_5 = ref_2_5.populations[pop_name].history
for i in range(len(hist_2_3)):
if hist_2_3[i] > 0:
ratio = hist_2_5[i] / hist_2_3[i]
assert np.abs(ratio - 1.) < 0.001
def test_point_estimate_skip_zeros():
start_day = 12
end_day = 60
ref_2 = Model.open_file(path_model_2_2)
sim_2 = Model.open_file(path_model_2_2)
# do fit of alpha_1, trans_rate_1_time
for par_name in ['alpha_1']:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
par = ref_2.parameters['trans_rate_1_time']
par.set_variable(None, None)
par.set_min(13)
par.set_max(19)
sim_2.reset()
rn_dict = sim_2.populations['reported'].get_report_noise()
rn_dict['report_days'].set_value(7)
sim_2.generate_data(end_day)
sim_2.populations['reported'].history[47] = np.inf
optimizer = Optimizer(ref_2,
'total reported',
sim_2.populations['reported'].history,
[start_day, end_day],
cumul_reset=True,
skip_data='42,45:48',
skip_zeros=True)
#optimizer = Optimizer(ref_2, 'total reported', sim_2.populations['reported'].history, [start_day, end_day],
# cumul_reset=True, skip_zeros=False)
optimizer.reset_variables()
scan_dict = optimizer.i_fit()
assert ref_2.parameters['trans_rate_1_time'].get_value() in [15, 16, 17]
par = ref_2.parameters['trans_rate_1_time']
par.set_fixed()
popt, pcov = optimizer.fit()
assert np.abs(ref_2.parameters['alpha_1'].get_value() -
ref_2.parameters['alpha_1'].initial_value) < 0.02
def test_model_2_8():
ref_2_8 = Model.open_file(path_model_2_8)
ref_2_8.transitions['outbreak_v'].enabled = True
ref_2_8.parameters['outbreak_v_time'].set_value(30)
ref_2_8.parameters['outbreak_v_number'].set_value(2.)
ref_2_8.parameters['alpha_0'].set_value(0.35)
# ref_2_8.reset()
# ref_2_8.evolve_expectations(200)
ref_2_8.reset()
ref_2_8.generate_data(200)
i = 1
def test_model_2_7():
ref_2_7 = Model.open_file(path_model_2_7)
ref_2_7.transitions['vaccination_1'].enabled = True
ref_2_7.transitions['vaccination_2'].enabled = True
ref_2_7.parameters['vacc_time_2'].set_value(80)
ref_2_7.parameters['vacc_number_2'].set_value(-5.)
# ref_2_7.reset()
# ref_2_7.evolve_expectations(200)
ref_2_7.reset()
ref_2_7.generate_data(200)
i = 1
def test_model_ks_4_2():
ks_4_2 = Model.open_file('ks_4_2_0116_problem.pypm')
ks_4_2.reset()
ks_4_2.evolve_expectations(720)
recent_oe_ve = [[ks_4_2.populations['os_susceptible'].history[i], ks_4_2.populations['ve_susceptible'].history[i]] \
for i in range(695,700)]
ks_4_2.reset()
ks_4_2.parameters['ve_frac'].set_value(0.801)
ks_4_2.evolve_expectations(720)
recent_oe_ve2 = [[ks_4_2.populations['os_susceptible'].history[i], ks_4_2.populations['ve_susceptible'].history[i]] \
for i in range(695,700)]
i = 1
def upload_model_files(self, model_filenames: list):
if not isinstance(model_filenames, list):
raise TypeError(
'Error in upload_model_files: argument must be a list of .pypm filenames'
)
for model_file in model_filenames:
if not isinstance(model_file, str):
raise TypeError(
'Error in upload_model_files: argument must be a list of .pypm filenames'
)
for model_filename in model_filenames:
self.__add_model(Model.open_file(model_filename))
def test_Model_copy_values_from():
ref_model_2 = Model.open_file(path_model_2)
alberta_model_2 = Model.open_file('ab_2_0514.pypm')
ref_model_2.copy_values_from(alberta_model_2)
ref_model_2.reset()
ref_model_2.evolve_expectations(80)
alberta_model_2.reset()
alberta_model_2.evolve_expectations(80)
for pop_name in ref_model_2.populations:
pop = ref_model_2.populations[pop_name]
# Bug fixed in ref_model_2, not in alberta_model_2 accounts for small differences in icu
if 'vent' not in pop.name and 'in_' not in pop.name:
ref_hist = pop.history
alberta_hist = alberta_model_2.populations[pop_name].history
for i in range(len(ref_hist)):
if np.abs(ref_hist[i] - alberta_hist[i]) > 0.0001:
print(pop.name, i, ref_hist[i], alberta_hist[i],
np.abs(ref_hist[i] - alberta_hist[i]))
assert np.abs(ref_hist[i] - alberta_hist[i]) < 0.001
def test_class_Population():
"""tests to ensure the behaviour class Population"""
init_value = 100
test_pop = Population('test population', init_value, description='For testing populations',
hidden=True, color='black', show_sim=False, report_noise=False,
report_noise_par=None)
assert test_pop.history[0] == init_value
model = Model.open_file(path_model_2_8)
test_pop.set_model(model)
incoming = 10
test_pop.update_future_fast(incoming)
assert test_pop.future[0] == incoming
test_pop.do_time_step()
assert test_pop.history[1] == init_value + incoming
assert len(test_pop.future) == 0 or test_pop.future == 0
scale_factor = 0.5
test_pop.scale_history(scale_factor)
assert test_pop.history[0] == init_value * scale_factor
assert test_pop.history[1] == (init_value + incoming) * scale_factor
# check that noise in reporting makes sense
# Expectations are not effected - data should be
# but the total reported should be the same after all reporting complete
noise_factor = Parameter('noise_factor', 0.3)
backlog_factor = Parameter('backlog_factor', 0.8)
# restart - back to initial value
for expectations in [True, False]:
test_pop.reset()
future = [0, 10, 100, 100, 100, 10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
future_sum = np.sum(np.array(future))
test_pop.set_report_noise(True, noise_factor, backlog_factor, None)
test_pop.future = future
for i in range(len(future) + 5):
test_pop.do_time_step(expectations=expectations)
assert test_pop.history[-1] == init_value + future_sum
report_days = Parameter('report_days', 127, parameter_min=-7, parameter_max=127, parameter_type='int')
for report_day_value in [127, 63, -1, -2, -5, -7]:
report_days.set_value(report_day_value)
test_pop.reset()
future = [0, 10, 100, 100, 100, 10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
future_sum = np.sum(np.array(future))
test_pop.set_report_noise(True, noise_factor, backlog_factor, report_days)
test_pop.future = future
for i in range(len(future) + 5):
test_pop.do_time_step(expectations=expectations)
assert test_pop.history[-1] == init_value + future_sum
def test_sim_gof_local():
start_day = 12
end_day = 60
ref_2 = Model.open_file(path_model_2_2)
sim_2 = Model.open_file(path_model_2_2)
# do fit of alpha_0, alpha_1, cont_0
par_names = ['alpha_1']
for par_name in par_names:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
sim_2.reset()
sim_2.generate_data(end_day)
sim_2.populations['reported'].history[47] = np.inf
optimizer = Optimizer(ref_2,
'total reported',
sim_2.populations['reported'].history,
[start_day, end_day],
cumul_reset=True,
skip_data='42,45:48')
optimizer.reset_variables()
popt, pcov = optimizer.fit()
fit_statistics = optimizer.fit_statistics
optimizer.calc_chi2s = False
optimizer.calc_chi2f = True
n_rep = 10
optimizer.calc_sim_gof(n_rep)
fit_stat_list = optimizer.fit_stat_list
ndof = fit_stat_list[0]['ndof']
chi2_list = [fit_stat_list[i]['chi2'] for i in range(n_rep)]
chi2_mean = np.mean(chi2_list)
assert np.abs(chi2_mean - ndof) < 8.E6
acor_list = [fit_stat_list[i]['acor'] for i in range(n_rep)]
acor_mean = np.mean(acor_list)
assert np.abs(acor_mean) < 0.2
def test_model_ny_4_2():
ny_4_2 = Model.open_file(
'/Users/karlen/pypm-local/models/covid19/USA/ny_4_2_0109.pypm')
ny_4_2.reset()
ny_4_2.evolve_expectations(40)
num = ny_4_2.populations['reported'].history[-1]
cont_0 = ny_4_2.parameters['cont_0'].get_value()
ny_4_2.parameters['cont_0'].set_value(2. * cont_0)
ny_4_2.reset()
ny_4_2.evolve_expectations(40)
num2 = ny_4_2.populations['reported'].history[-1]
if (ny_4_2.populations['contagious_o'].initial_value !=
ny_4_2.parameters['cont_0']):
ny_4_2.populations['contagious_o'].initial_value = ny_4_2.parameters[
'cont_0']
print('changed')
ny_4_2.reset()
ny_4_2.evolve_expectations(40)
num3 = ny_4_2.populations['reported'].history[-1]
i = 1
def test_linear_modifier():
ref_2_4 = Model.open_file(path_model_2_4)
trajectory = Trajectory(ref_2_4, 'contagious', 'trans_rate_1', [0.02, 2.0])
alpha_c = trajectory.get_alpha(0.)
ref_2_4.parameters['alpha_1'].set_value(alpha_c)
ref_2_4.parameters['to_icu_delay_mean'].set_value(0.5)
ref_2_4.parameters['to_icu_delay_sigma'].set_value(0.5)
ref_2_4.transitions['mod_icu_frac'].enabled = True
ref_2_4.parameters['icu_frac_time'].set_value(60)
ref_2_4.parameters['icu_frac_0'].set_value(0.1)
# the meaning of this parameter changed to the end_value
# ref_2_4.parameters['icu_frac_slope'].set_value(0.01)
ref_2_4.parameters['icu_frac_slope'].set_value(0.2)
ref_2_4.parameters['icu_frac_nstep'].set_value(10)
ref_2_4.reset()
ref_2_4.evolve_expectations(100)
pop_hist = ref_2_4.populations['icu admissions'].history
assert np.abs(pop_hist[55] - pop_hist[54] - pop_hist[50] +
pop_hist[49]) < 0.01
assert np.abs(pop_hist[75] - pop_hist[74] - pop_hist[70] +
pop_hist[69]) < 0.1
assert np.abs((pop_hist[75] - pop_hist[74]) /
(pop_hist[55] - pop_hist[54]) - 2.) < 0.1
def test_mixing_evolve_data():
ref_2_9 = Model.open_file(path_model_2_9)
ref_2_9.reset()
ref_2_9.generate_data(100, from_step=0, data_start=50)
reported = ref_2_9.populations['reported'].history
i = 1
def test_model_az_2_9():
az_2_8 = Model.open_file('az_2_8_0530.pypm')
az_2_9 = Model.open_file(path_model_2_9)
az_2_9.copy_values_from(az_2_8)
az_2_9.name = 'az_2_9_0530'
#v_hesitant = 0.14
#N0 = az_2_8.parameters['N_0'].get_value()
#az_2_8.populations['vacc cand'].set_initial_value(int(N0 * (1. - v_hesitant)))
#az_2_8.populations['sus vacc cand'].set_initial_value(int(N0 * (1. - v_hesitant)))
#az_2_8.parameters['vacc_time_1'].set_value(290)
#az_2_9.parameters['vacc_time_1'].set_value(290)
az_2_8.reset()
az_2_8.evolve_expectations(289)
az_2_8.evolve_expectations(500 - 289, from_step=289)
az_2_9.parameters['vac_waned_delay_mean'].set_value(10000.)
az_2_9.parameters['nat_waned_delay_mean'].set_value(10000.)
az_2_9.parameters['vac_waned_frac'].set_value(0.)
#v_hesitant = 0.14
#N0 = az_2_9.parameters['N_0'].get_value()
#az_2_9.populations['vacc cand'].set_initial_value(int(N0 * (1. - v_hesitant)))
#az_2_9.populations['sus vacc cand'].set_initial_value(int(N0 * (1. - v_hesitant)))
az_2_9.reset()
#az_2_9.evolve_expectations(500)
az_2_9.evolve_expectations(289)
az_2_9.evolve_expectations(500 - 289, from_step=289)
reps = []
rep_vs = []
wanned = []
pops = [
'susceptible', 'immunized', 'usefully vaccinated', 'daily vaccinated',
'vacc cand', 'sus vacc cand'
]
compares = {}
for pop in pops:
compares[pop] = []
for day in [288, 289, 290, 291, 292, 293, 500]:
reps_m = []
rep_vs_m = []
for model in [az_2_8, az_2_9]:
rep = model.populations['reported'].history[
day] - model.populations['reported'].history[day - 1]
reps_m.append(rep)
rep_v = model.populations['reported_v'].history[
day] - model.populations['reported_v'].history[day - 1]
rep_vs_m.append(rep_v)
for pop in pops:
compares[pop].append([
az_2_8.populations[pop].history[day],
az_2_9.populations[pop].history[day]
])
wanned.append(az_2_9.populations['waned_immunity'].history[day])
reps.append(reps_m)
rep_vs.append(rep_vs_m)
i = 1
def get_csv(self, forecast_date, category,cor_scale=1.):
t0 = datetime.date(2020, 3, 1)
forecast_date_text = forecast_date.isoformat()
day_of_week = forecast_date.weekday()
days_after_t0 = (forecast_date - t0).days
first_sunday = days_after_t0
if day_of_week == 0:
first_sunday -= 1
elif day_of_week < 6:
first_sunday += 6 - day_of_week
#names = ['case', 'death']
#dict_names = ['case', 'death']
#periods = ['wk', 'wk']
#inc_types_list = [['inc', 'cum'], ['inc', 'cum']]
period = 'wk'
inc_types = ['inc', 'cum']
# collect information for all Germany
de_inc_point_est_dict = {}
de_inc_periods_dict = {}
de_cum_point_est_dict = {}
de_cum_periods_dict = {}
de_inc_point_est_dict[category] = {}
de_inc_periods_dict[category] = {}
de_cum_point_est_dict[category] = {}
de_cum_periods_dict[category] = {}
state_sums = 0
for state in self.regional_abbreviations:
abbrev = self.regional_abbreviations[state]
location = self.regional_locations[state]
state_data = None
if category == 'case':
state_data = self.pd_dict['germany-rki-pypm.csv'][abbrev + '-pt'].fillna(0).values
elif category == 'death':
state_data = self.pd_dict['germany-rki-pypm.csv'][abbrev + '-dt'].fillna(0).values
success = False
model = None
if category == 'case':
for model_name in self.model_names:
try:
filename = abbrev + model_name + '.pypm'
path_model = self.model_dir / filename
model = Model.open_file(path_model)
success = True
break
except:
pass
elif category == 'death':
for model_name in self.model_names:
for model_suffix in ['_d','_h','']:
try:
filename = abbrev + model_name + model_suffix + '.pypm'
path_model = self.model_dir / filename
model = Model.open_file(path_model)
success = True
break
except:
pass
if not success:
raise RuntimeError('No model for: ', abbrev)
print(model.name)
hub_dict = model.user_dict['forecast_hub'][forecast_date][category]
point_est_dict = hub_dict['point_estimates']
quantile_dict = hub_dict['quantiles']
inc_periods = hub_dict['inc_periods']
cum_periods = [state_data[first_sunday-1]] * len(inc_periods[0])
for inc_type in inc_types:
# for cum record
sum = state_data[first_sunday-1]
if inc_type == 'cum':
state_sums += sum
for i_period in point_est_dict:
index = int(i_period) - 1
target = i_period+' '+period+' ahead '+inc_type+' '+category
days = int(i_period)
if period == 'wk':
days *= 7
else:
days += 1
target_end_date = (t0 + timedelta(days=(first_sunday + days - 1))).isoformat()
# point estimates
value = point_est_dict[i_period]
if inc_type == 'cum':
sum += point_est_dict[i_period]
value = sum
if i_period in de_cum_point_est_dict[category]:
de_cum_point_est_dict[category][i_period] += sum
else:
de_cum_point_est_dict[category][i_period] = sum
self.add_record(forecast_date_text,target,target_end_date,location,'point','NA',value)
# quantiles
if inc_type == 'inc':
if i_period in de_inc_point_est_dict[category]:
de_inc_point_est_dict[category][i_period] += value
else:
de_inc_point_est_dict[category][i_period] = value
for quant in quantile_dict[i_period]:
value = quantile_dict[i_period][quant]
self.add_record(forecast_date_text, target, target_end_date, location, 'quantile', quant, value)
if i_period in de_inc_periods_dict[category]:
ip_len = len(inc_periods[index])
for i_rep in range(len(de_inc_periods_dict[category][i_period])):
# in case there are fewer repetitions, loop over others:
ip_index = i_rep % ip_len
de_inc_periods_dict[category][i_period][i_rep] += inc_periods[index][ip_index]
else:
de_inc_periods_dict[category][i_period] = copy.copy(inc_periods[index])
elif inc_type == 'cum':
for i_rep in range(len(inc_periods[index])):
cum_periods[i_rep] += inc_periods[index][i_rep]
for quant in quantile_dict[i_period]:
value = np.percentile(cum_periods, float(quant)*100.)
self.add_record(forecast_date_text, target, target_end_date, location, 'quantile', quant, value)
if i_period in de_cum_periods_dict[category]:
ip_len = len(cum_periods)
for i_rep in range(len(de_cum_periods_dict[category][i_period])):
# in case there are fewer repetitions, loop over others:
ip_index = i_rep % ip_len
de_cum_periods_dict[category][i_period][i_rep] += cum_periods[ip_index]
else:
de_cum_periods_dict[category][i_period] = [cum_periods[i_rep] for i_rep in range(len(cum_periods))]
# return 'Germany' summary:
# there is a separate file with Germany wide deaths. Need to correct by adding the additional deaths here
#additional_deaths = de_deaths - state_deaths
additional_counts = 0
if category == 'death':
print('Germany total deaths included:',state_sums)
elif category == 'case':
print('Germany total cases included:', state_sums)
#print('Germany total: additional deaths included:', additional_deaths)
location = self.regional_locations['Germany']
quants = [0.01, 0.025] + [0.05 + 0.05 * i for i in range(19)] + [0.975, 0.99]
#for i in range(2):
for inc_type in inc_types:
for i_period in de_inc_point_est_dict[category]:
target = i_period + ' ' + period + ' ahead ' + inc_type + ' ' + category
days = int(i_period)
if period == 'wk':
days *= 7
else:
days += 1
target_end_date = (t0 + timedelta(days=(first_sunday + days - 1))).isoformat()
# point estimates
value = de_inc_point_est_dict[category][i_period]
if inc_type == 'cum':
value = de_cum_point_est_dict[category][i_period] + additional_counts
self.add_record(forecast_date_text, target, target_end_date, location, 'point', 'NA', value)
# quantiles
if inc_type == 'inc':
median = np.percentile(de_inc_periods_dict[category][i_period], 50.)
for quant in quants:
value = np.percentile(de_inc_periods_dict[category][i_period], float(quant)*100.)
# increase the width of the interval for the US as a whole, to account for correlations between states
scaled_value = max(0., cor_scale * (value - median) + median)
quant_text = '{0:0.3f}'.format(quant)
self.add_record(forecast_date_text, target, target_end_date, location, 'quantile', quant_text,
scaled_value)
elif inc_type == 'cum':
median = np.percentile(de_cum_periods_dict[category][i_period], 50.) + additional_counts
for quant in quants:
value = np.percentile(de_cum_periods_dict[category][i_period], float(quant) * 100.) + additional_counts
scaled_value = max(0., cor_scale * (value - median) + median)
quant_text = '{0:0.3f}'.format(quant)
self.add_record(forecast_date_text, target, target_end_date, location, 'quantile', quant_text,
scaled_value)
return self.buff
def __init__(self, name: str, model):
"""Constructor
- Model specifies the reference model: populations etc should match those in
the ensemble of models. It can be specified as a model, or as a filename
"""
super().__init__(name)
self.model = None
in_model = None
if isinstance(model, Model):
in_model = model
elif isinstance(model, str):
in_model = Model.open_file(model)
else:
raise TypeError(
'Error creating ensemble. The model argument needs to be a model object '
+ 'or have a .pypm filename')
# to avoid the possibility that the reference model is subsequently modified, make a copy
# that will not be as exposed to tampering
self.model = copy.deepcopy(in_model)
# point to the population objects in self.model to save the histories
# this allows for consistent access to history between models and ensembles
for pop_name in self.model.populations:
pop = self.model.populations[pop_name]
self.populations[pop_name] = pop
# the parameters will be set once the list of models are read in
self.parameters = {}
# the boot parameters point to the ones copied from the reference
self.boot_pars = self.model.boot_pars
# the important scaling parameter for the ensemble gets defined here
# this is forced to be a parameter when each model is created and
# checked when the models are included in the ensemble
boot_pop = self.populations[self.boot_pars['boot_population']]
boot_pop_ip = boot_pop.initial_value
self.parameters[boot_pop_ip.name] = boot_pop_ip
# When this scaling parameter is changed, all the model scaling parameters
# must also be scaled accordingly.
boot_pop_ip.set_must_update(self)
# The list of models (either provided directly or as a list of .pypm files)
# They are required to have unique names for user interaction
# Ordering is preserved to match the transmission matrix
self.model_list = []
self.models = {}
self.infected_name = None
self.susceptible_name = None
self.total_name = None
self.contagious_name = None
self.alpha_name = None
self.infection_cycle_name = None
self.contact = None
self.contact_matrix = None
self.contact_type = ''
self.null_pop = Population('null', 0.)
self.total_goal = None
self.boot_achieved = {}
self.max_scale_factor = None
self.__distribution = None
self.__nbinom_par = None
self.set_distribution('poisson', None)
def test_model_bc_4_1():
bc_4_1 = Model.open_file('bc_4_1_1231x.pypm')
bc_4_1.reset()
bc_4_1.evolve_expectations(700)
i = 1
def test_Ensemble_properties_equality():
"""tests to ensure the properties of Ensemble"""
# Test that the ensemble of two identical models, with different scales
# and equality contact matrix behaves like the same ensemble
# with independent contact matrix
test_a1 = Model.open_file(path_model_2_3)
test_a1.name = 'test_a1'
test_b1 = Model.open_file(path_model_2_3)
test_b1.name = 'test_b1'
reference1 = Model.open_file(path_model_2_3)
reference1.name = 'reference1'
cont_0 = test_b1.parameters['cont_0'].get_value()
test_b1.parameters['cont_0'].set_value(cont_0 * 2.)
N_0 = test_b1.parameters['N_0'].get_value()
test_b1.parameters['N_0'].set_value(N_0 * 2)
test_ensemble1 = Ensemble('test_ensemble1', reference1)
test_ensemble1.upload_models([test_a1, test_b1])
test_ensemble1.define_cross_transmission('infection cycle',
'infected',
'susceptible',
'total',
'contagious',
'alpha',
contact_type='diagonal')
test_a2 = Model.open_file(path_model_2_3)
test_a2.name = 'test_a2'
test_b2 = Model.open_file(path_model_2_3)
test_b2.name = 'test_b2'
reference2 = Model.open_file(path_model_2_3)
reference2.name = 'reference2'
cont_0 = test_b2.parameters['cont_0'].get_value()
test_b2.parameters['cont_0'].set_value(cont_0 * 2.)
N_0 = test_b2.parameters['N_0'].get_value()
test_b2.parameters['N_0'].set_value(N_0 * 2)
test_ensemble2 = Ensemble('test_ensemble2', reference2)
test_ensemble2.upload_models([test_a2, test_b2])
test_ensemble2.define_cross_transmission('infection cycle',
'infected',
'susceptible',
'total',
'contagious',
'alpha',
contact_type='equality')
n_days = 100
test_ensemble1.reset()
test_ensemble1.evolve_expectations(n_days)
test_ensemble2.reset()
test_ensemble2.evolve_expectations(n_days)
for pop_name in test_ensemble1.populations:
pop1 = test_ensemble1.populations[pop_name]
if pop1.show_sim:
pop1h = test_ensemble1.populations[pop_name].history
pop2h = test_ensemble2.populations[pop_name].history
for i in range(len(pop1h)):
if pop2h[i] > 0.:
ratio = pop1h[i] / pop2h[i]
assert np.abs(ratio - 1.) < 0.03
def test_Model_copy_values_from(self):
path_model_2 = self.model_dir / 'ref_model_2.pypm'
ref_model_2 = Model.open_file(path_model_2)
def test_report_noise_days():
for report_noise_weekly in [False, True]:
start_day = 12
end_day = 80
ref_2 = Model.open_file(path_model_2_2)
ref_2.parameters['report_noise'].set_value(0.1)
# BC: no reporting on Sundays
ref_2.parameters['report_days'].set_value(63)
sim_2 = copy.deepcopy(ref_2)
sim_2.populations['reported'].report_noise_weekly = report_noise_weekly
# do fit of alpha_0, alpha_1, cont_0
par_names = ['alpha_0', 'alpha_1', 'cont_0']
sums = {}
sum2s = {}
for par_name in par_names:
par = ref_2.parameters[par_name]
par.set_variable(None, None)
sums[par_name] = 0.
sum2s[par_name] = 0.
n_rep = 10
fit_stat_list = []
for i in range(n_rep):
sim_2.reset()
sim_2.generate_data(end_day)
optimizer = Optimizer(ref_2, 'total reported',
sim_2.populations['reported'].history,
[start_day, end_day])
optimizer.reset_variables()
popt, pcov = optimizer.fit()
fit_stat_list.append(optimizer.fit_statistics)
for par_name in par_names:
value = ref_2.parameters[par_name].get_value()
sums[par_name] += value
sum2s[par_name] += value**2
ass_std = {}
ass_std['alpha_0'] = 0.05
ass_std['alpha_1'] = 0.01
ass_std['cont_0'] = 10.
means = {}
std = {}
for par_name in par_names:
means[par_name] = sums[par_name] / n_rep
std[par_name] = np.sqrt(sum2s[par_name] / n_rep -
means[par_name]**2)
for par_name in par_names:
assert std[par_name] < ass_std[par_name]
truth = ref_2.parameters[par_name].initial_value
assert np.abs((means[par_name] - truth) / std[par_name] /
np.sqrt(1. * n_rep)) < 3.
ndof = fit_stat_list[0]['ndof']
chi2_list = [fit_stat_list[i]['chi2'] for i in range(n_rep)]
chi2_mean = np.mean(chi2_list)
assert chi2_mean / ndof > 2.5
if not report_noise_weekly:
acor_list = [fit_stat_list[i]['acor'] for i in range(n_rep)]
acor_mean = np.mean(acor_list)
assert acor_mean < -0.2
