def implementation_checks(window: int = 7):
df = get_dataframe()
ts = get_time_series(df)
I = ts.Hospitalized.values
for kernel in kernels.keys():
plt.figure()
plt.bar(x = ts.index, height = I, label = "raw")
smoother = convolution(kernel, window)
plt.plot(ts.index, smoother(I)[:-window+1], label = "smooth")
plt.title(f"kernel: {kernel} (window: {window})", loc = "left")
plt.show()
if __name__ == "__main__":
root = cwd()
data = root / "data"
figs = root / "figs"
gamma = 0.2
window = 2
num_migrants = 600000
release_rate = 0.5
state_cases = etl.load_cases(data / "Bihar_Case_data_May20.csv")
district_cases = etl.split_cases_by_district(state_cases)
district_ts = {
district: etl.get_time_series(cases)
for (district, cases) in district_cases.items()
}
R_mandatory = {
district: estimate(district, ts, use_last=True)
for (district, ts) in district_ts.items()
}
districts, pops, migrations = etl.district_migration_matrix(
data / "Migration Matrix - District.csv")
for district in districts:
if district not in R_mandatory.keys():
R_mandatory[district] = 1.5
R_voluntary = {district: 1.5 * R for (district, R) in R_mandatory.items()}
migration_spike = etl.migratory_influx_matrix(
data = Path("./data")
figs = Path("./figs/comparison/original")
# define data versions for api files
paths = {
"v3": ["raw_data1.csv", "raw_data2.csv"],
"v4": ["raw_data3.csv", "raw_data4.csv", "raw_data5.csv", "raw_data6.csv", "raw_data7.csv"]
}
df = load_all_data(
v3_paths = [data/filepath for filepath in paths['v3']],
v4_paths = [data/filepath for filepath in paths['v4']]
)
ts = get_time_series(df, "detected_state")
for state in states:
from scipy.stats import gamma # not sure why this needs to be recalled after each state, but otherwite get a type exception
fig, ax = plt.subplots()
# g=open('covid19-in-india-2/covid_19_india.csv', 'r')
# reader=csv.reader(g)
# confirmed=[]
# dead=[]
# recovered=[]
# dates=[]
# day=[]
}
# define data versions for api files
paths = {
"v3": ["raw_data1.csv", "raw_data2.csv"],
"v4": ["raw_data3.csv", "raw_data4.csv"]
}
# download data from india covid 19 api
for f in paths['v3'] + paths['v4']:
download_data(data, f)
# run rolling regressions on historical national case data
dfn = load_all_data(v3_paths=paths['v3'], v4_paths=paths['v4'])
data_recency = str(dfn["date_announced"].max()).split()[0]
tsn = get_time_series(dfn)
grn = run_regressions(tsn, window=5, infectious_period=1 / gamma)
# disaggregate down to states
dfs = {state: dfn[dfn["detected_state"] == state] for state in states}
tss = {state: get_time_series(cases) for (state, cases) in dfs.items()}
for (_, ts) in tss.items():
ts['Hospitalized'] *= prevalence
grs = {
state: run_regressions(timeseries,
window=5,
infectious_period=1 / gamma)
for (state, timeseries) in tss.items() if len(timeseries) > 5
}
# voluntary and mandatory reproductive numbers
def project(p: pd.Series):
t = (p.R - p.Intercept)/p.gradient
return (max(0, p.R), max(0, p.Intercept + p.gradient*(t + 7)), max(0, p.Intercept + p.gradient*(t + 14)), np.sqrt(p.gradient_stderr))
if __name__ == "__main__":
root = cwd()
data = root/"data"
figs = root/"figs"
gamma = 0.2
window = 5
state_cases = etl.load_cases(data/"Bihar_Case_data_May20.csv")
district_cases = etl.split_cases_by_district(state_cases)
district_ts = {district: etl.get_time_series(cases) for (district, cases) in district_cases.items()}
R_mandatory = {district: estimate(district, ts, use_last = True) for (district, ts) in district_ts.items()}
districts, pops, migrations = etl.district_migration_matrix(data/"Migration Matrix - District.csv")
for district in districts:
if district not in R_mandatory.keys():
R_mandatory[district] = 1.5
R_voluntary = {district: 1.5*R for (district, R) in R_mandatory.items()}
si, sf = 0, 1000
simulation_results = [
run_policies(district_ts, pops, districts, migrations, gamma, R_mandatory, R_voluntary, seed = seed)
for seed in tqdm(range(si, sf))
]
np.sqrt(p.gradient_stderr))
if __name__ == "__main__":
root = cwd()
data = root / "data"
figs = root / "figs"
gamma = 0.2
window = 5
state_cases = etl.load_cases(data / "Bihar_Case_data_May11.csv")
state_cases14 = etl.load_cases(data / "Bihar_Case_data_May14.csv")
district_cases = etl.split_cases_by_district(state_cases)
district_ts = {
district: etl.get_time_series(cases)
for (district, cases) in district_cases.items()
}
R_mandatory = {
district: estimate(district, ts, window=window, use_last=True)
for (district, ts) in district_ts.items()
}
districts, pops, migrations = etl.district_migration_matrix(
data / "Migration Matrix - District.csv")
for district in districts:
if district not in R_mandatory.keys():
R_mandatory[district] = 1.5
R_voluntary = {district: 1.5 * R for (district, R) in R_mandatory.items()}
si, sf = 0, 1000
def state_checks():
root = cwd()
data = root/"data"
# model details
gamma = 0.2
prevalence = 1
dfn = get_dataframe()
states = ["Maharashtra", "Andhra Pradesh", "Tamil Nadu", "Madhya Pradesh", "Punjab", "Gujarat", "Kerala", "Bihar"]
# first, check reimplementation against known figures
CI = 0.99
for state in states:
ts = get_time_series(dfn[
(dfn.date_announced <= "2020-05-08") &
(dfn.detected_state == state)
])
(
dates,
RR_pred, RR_CI_upper, RR_CI_lower,
T_pred, T_CI_upper, T_CI_lower,
total_cases, new_cases_ts,
anomalies, anomaly_date
) = analytical_MPVS(ts.Hospitalized, CI = CI)
plot_RR_est(dates, RR_pred, RR_CI_upper, RR_CI_lower, CI)
plt.title(f"{state}: Estimated $R_t$", loc="left", fontsize=20)
plt.figure()
plot_T_anomalies(dates, T_pred, T_CI_upper, T_CI_lower, new_cases_ts, anomaly_dates, anomalies, CI)
plt.title(f"{state}: Predicted Cases", loc="left", fontsize=20)
plt.show()
# generate smoothing checks for national time series
ts = get_time_series(dfn)
CI = 0.95
# 1: box filter
RRfig, RRax = plt.subplots(3, 1, sharex=True)
T_fig, T_ax = plt.subplots(3, 1, sharex=True)
for (i, n) in enumerate((5, 10, 15)):
(
dates,
RR_pred, RR_CI_upper, RR_CI_lower,
T_pred, T_CI_upper, T_CI_lower,
total_cases, new_cases_ts,
anomalies, anomaly_dates
) = analytical_MPVS(ts.Hospitalized, CI = CI, smoothing = lambda ts: box_filter(ts, n, None))
plt.sca(RRax[i])
plot_RR_est(dates, RR_pred, RR_CI_upper, RR_CI_lower, CI)
plt.title(f"smoothing window: {n}, no local smoothing", loc = "left", fontsize = 16)
plt.sca(T_ax[i])
plot_T_anomalies(dates, T_pred, T_CI_upper, T_CI_lower, new_cases_ts, anomaly_dates, anomalies, CI)
plt.title(f"smoothing window: {n}, no local smoothing", loc = "left", fontsize = 16)
plt.figure(RRfig.number)
RRfig.tight_layout()
plt.suptitle("Estimated Rt (box filter convolution)", fontsize = 20)
plt.figure(T_fig.number)
T_fig.tight_layout()
plt.suptitle("Estimated Daily Cases (box filter convolution)", fontsize = 20)
plt.show()
# 2: box filter, local smooothing
RRfig, RRax = plt.subplots(3, 1, sharex=True)
T_fig, T_ax = plt.subplots(3, 1, sharex=True)
for (i, n) in enumerate((5, 10, 15)):
s = n//2 + 1
(
dates,
RR_pred, RR_CI_upper, RR_CI_lower,
T_pred, T_CI_upper, T_CI_lower,
total_cases, new_cases_ts,
anomalies, anomaly_dates
) = analytical_MPVS(ts.Hospitalized, CI = CI, smoothing = lambda ts: box_filter(ts, n, s))
plt.sca(RRax[i])
plot_RR_est(dates, RR_pred, RR_CI_upper, RR_CI_lower, CI)
plt.title(f"smoothing window: {n}, local smoothing: last {s} points", loc = "left", fontsize = 16)
plt.sca(T_ax[i])
plot_T_anomalies(dates, T_pred, T_CI_upper, T_CI_lower, new_cases_ts, anomaly_dates, anomalies, CI)
plt.title(f"smoothing window: {n}, local smoothing: last {s} points", loc = "left", fontsize = 16)
plt.figure(RRfig.number)
RRfig.tight_layout()
plt.suptitle("Estimated Rt (box filter convolution, local smoothing)", fontsize = 20)
plt.subplots_adjust(right = 0.9)
plt.figure(T_fig.number)
T_fig.tight_layout()
plt.suptitle("Estimated Daily Cases (box filter convolution, local smoothing)", fontsize = 20)
plt.subplots_adjust(right = 0.9)
plt.show()