def __init__(self, fips, model_ensemble, county_outputs, filename, summary, xlim=(0, 360)):
self.fips = fips
self.county_outputs = county_outputs
self.model_ensemble = model_ensemble
self.county_metadata = load_data.load_county_metadata_by_fips(fips)
self.summary = summary
timeseries = combined_datasets.get_timeseries_for_fips(
fips, columns=[CommonFields.CASES, CommonFields.DEATHS], min_range_with_some_value=True
)
self.county_case_data = timeseries.data
self.report = PDFReport(filename=filename)
self.xlim = xlim
def __init__(self,
fips,
model_ensemble,
county_outputs,
filename,
summary,
xlim=(0, 360)):
self.fips = fips
self.county_outputs = county_outputs
self.model_ensemble = model_ensemble
self.county_metadata = load_data.load_county_metadata_by_fips(fips)
self.summary = summary
_county_case_data = load_data.load_county_case_data()
self.county_case_data = _county_case_data[_county_case_data['fips'] ==
fips]
self.report = PDFReport(filename=filename)
self.xlim = xlim
def generate_report(self):
"""
Generate a full report for the state.
"""
report = PDFReport(filename=self.filename)
for compartment in self.plot_compartments:
fig = self.plot_compartment(compartment)
report.add_figure(fig=fig)
report.close()
self.generate_surge_spreadsheet()
class CountyReport:
"""
Generate a county level report showing detailed data on different scenarios.
Parameters
----------
fips: str
County fips code.
model_ensemble: list(SEIRModel)
Models from the ensemble.
county_outputs: dict
Dictionary of structure (in yaml syntax)
suppression_policy__0.5:
t_list: array of timestamps
HICU: For compartment S we have a bunch of properties
S_ci50: the median posterior model. Other condidence intervals are available.
...
peak_value_ciXX:
peak_time_ciXX:
...
# For HICU, HVent, HGen, capacities are relevant, so we also calculate surge windows.
capacity: estimated capacity for this compartment.
surge_start: array surge window starts for each model in the ensemble
surge_end:
filename: str
Where to save the report.
summary: dict
Summary attrs to print on the first page of the report.
xlim: tuple
Limits in days since simulation start to plot.
"""
def __init__(self,
fips,
model_ensemble,
county_outputs,
filename,
summary,
xlim=(0, 360)):
self.fips = fips
self.county_outputs = county_outputs
self.model_ensemble = model_ensemble
self.county_metadata = load_data.load_county_metadata_by_fips(fips)
self.summary = summary
_county_case_data = load_data.load_county_case_data()
self.county_case_data = _county_case_data[_county_case_data['fips'] ==
fips]
self.report = PDFReport(filename=filename)
self.xlim = xlim
def generate_and_save(self):
"""
Name says it all.
"""
self.write_header_pages()
self.plot_seir_distributions()
self.report.close()
def write_header_pages(self):
"""
Generate header pages with the county summary and PDF summaries.
"""
self.report.write_text_page(
self.summary,
title=
f'PySEIR COVID19 Estimates\n{self.county_metadata["county"]} County, {self.county_metadata["state"]}',
page_heading=f'Generated {self.summary["date_generated"]}',
body_fontsize=6,
title_fontsize=12)
self.report.write_text_page(inspect.getsource(
ParameterEnsembleGenerator.sample_seir_parameters),
title='PySEIR Model Ensemble Parameters')
def plot_seir_distributions(self, xlim=(0, 360)):
"""
Generate plots for each suppression policy containing distributions and
peak information for each model compartment.
Parameters
----------
xlim: tuple
Limits in days since simulation start to plot.
"""
# Add a sample model from the ensemble.
fig = self.model_ensemble[0].plot_results(xlim=xlim)
fig.suptitle(
f'PySEIR COVID19 Estimates: {self.county_metadata["county"]} County, {self.county_metadata["state"]}. '
f'SAMPLE OF MODEL ENSEMBLE',
fontsize=16)
self.report.add_figure(fig)
for suppression_policy, output in self.county_outputs.items():
compartments = list(output.keys())
compartments.remove('t_list')
# -----------------------------------
# Plot each compartment distribution
# -----------------------------------
fig = plt.figure(figsize=(20, 24))
fig.suptitle(
f'PySEIR COVID19 Estimates: {self.county_metadata["county"]} County, {self.county_metadata["state"]}. '
f'\nSupression Policy={suppression_policy} (1=No Suppression)',
fontsize=16)
for i_plot, compartment in enumerate(compartments):
plt.subplot(5, 5, i_plot + 1)
plt.plot(output['t_list'],
output[compartment]['ci_50'],
color='steelblue',
linewidth=3,
label=compartment_to_name_map[compartment])
plt.fill_between(output['t_list'],
output[compartment]['ci_32'],
output[compartment]['ci_68'],
alpha=.3,
color='steelblue')
plt.fill_between(output['t_list'],
output[compartment]['ci_5'],
output[compartment]['ci_95'],
alpha=.3,
color='steelblue')
plt.yscale('log')
plt.ylim(1e0)
plt.xlim(0, 360)
plt.grid(True, which='both', alpha=0.3)
plt.xlabel('Days Since Case 0')
# Circular import :(
from pyseir.ensembles.ensemble_runner import compartment_to_capacity_attr_map
if compartment in compartment_to_capacity_attr_map:
percentiles = np.percentile([
getattr(m,
compartment_to_capacity_attr_map[compartment])
for m in self.model_ensemble
], (5, 32, 50, 68, 95))
plt.hlines(percentiles[2],
*plt.xlim(),
label='ICU Capacity',
color='darkseagreen')
plt.hlines([percentiles[0], percentiles[4]],
*plt.xlim(),
color='darkseagreen',
linestyles='-.',
alpha=.4)
plt.hlines([percentiles[1], percentiles[3]],
*plt.xlim(),
color='darkseagreen',
linestyles='--',
alpha=.2)
# Plot data
if compartment in ['D', 'total_deaths'
] and len(self.county_case_data) > 0:
plt.errorbar((self.county_case_data.date -
self.summary['t0']).dt.days,
self.county_case_data.deaths,
yerr=np.sqrt(self.county_case_data.deaths),
linestyle='-',
label='Deaths Observed',
marker='o',
markersize=4)
if compartment in ['I'] and len(self.county_case_data) > 0:
plt.errorbar((self.county_case_data.date -
self.summary['t0']).dt.days,
self.county_case_data.cases,
yerr=np.sqrt(self.county_case_data.cases),
linestyle='-',
label='Cases Observed',
marker='o',
markersize=4,
color='firebrick')
plt.legend()
self._plot_dates(log=False)
# -----------------------------
# Plot peak Timing
# -----------------------------
color_cycle = plt.rcParams['axes.prop_cycle'].by_key(
)['color'] + list('bgrcmyk')
marker_cycle = ['o', 's', '+', 'd', 'o'] * 4
plt.subplot(5, 5, len(compartments) + 1)
for i, compartment in enumerate([
'E', 'A', 'I', 'HGen', 'HICU', 'HVent',
'general_admissions_per_day', 'icu_admissions_per_day',
'direct_deaths_per_day', 'total_deaths_per_day'
]):
median = output[compartment]['peak_time_ci50']
ci5, ci95 = output[compartment]['peak_time_ci5'], output[
compartment]['peak_time_ci95']
ci32, ci68 = output[compartment]['peak_time_ci32'], output[
compartment]['peak_time_ci68']
plt.scatter(median,
i,
label=compartment_to_name_map[compartment],
c=color_cycle[i],
marker=marker_cycle[i])
plt.fill_betweenx(
[i - .3, i + .3],
[ci32, ci32],
[ci68, ci68],
alpha=.3,
color=color_cycle[i],
)
plt.fill_betweenx([i - .1, i + .1], [ci5, ci5], [ci95, ci95],
alpha=.3,
color=color_cycle[i])
self._plot_dates(log=False)
plt.legend(loc=(1.05, 0.0))
plt.grid(True, which='both', alpha=0.3)
plt.xlabel('Peak Time After $t_0(C=5)$ [Days]')
plt.yticks([])
# -----------------------------
# Plot peak capacity
# -----------------------------
plt.subplot(5, 5, len(compartments) + 3)
for i, compartment in enumerate([
'E', 'A', 'I', 'R', 'D', 'total_deaths',
'direct_deaths_per_day', 'total_deaths_per_day', 'HGen',
'HICU', 'HVent', 'HGen_cumulative', 'HICU_cumulative',
'HVent_cumulative', 'general_admissions_per_day',
'icu_admissions_per_day'
]):
median = output[compartment]['peak_value_ci50']
ci5, ci95 = output[compartment]['peak_value_ci5'], output[
compartment]['peak_value_ci95']
ci32, ci68 = output[compartment]['peak_value_ci32'], output[
compartment]['peak_value_ci68']
plt.scatter(median,
i,
label=compartment_to_name_map[compartment],
c=color_cycle[i],
marker=marker_cycle[i])
plt.fill_betweenx([i - .3, i + .3], [ci32, ci32], [ci68, ci68],
alpha=.3,
color=color_cycle[i])
plt.fill_betweenx([i - .1, i + .1], [ci5, ci5], [ci95, ci95],
alpha=.3,
color=color_cycle[i])
plt.xscale('log')
plt.vlines(self.county_metadata['total_population'],
*plt.ylim(),
label='Entire Population',
alpha=0.5,
color='g')
plt.vlines(self.county_metadata['total_population'] * 0.65,
*plt.ylim(),
label='Approx. Herd Immunity',
alpha=0.5,
color='purple',
linestyles='--',
linewidths=2)
plt.legend(loc=(1, -0.1))
plt.grid(True, which='both', alpha=0.3)
plt.xlabel('Value at Peak')
plt.yticks([])
self.report.add_figure(fig)
def _plot_dates(self, log=True):
"""
Helper function to add date plots.
Parameters
----------
log: bool
If True, shift y-positioning of labels based on a log scale.
"""
low_limit = plt.ylim()[0]
if log:
upp_limit = 1 * np.log(plt.ylim()[1])
else:
upp_limit = 1 * plt.ylim()[1]
for month in range(4, 11):
dt = datetime.datetime(day=1, month=month, year=2020)
offset = (dt - self.summary['t0']).days
plt.vlines(offset,
low_limit,
upp_limit,
color='firebrick',
alpha=.4,
linestyles=':')
plt.text(offset,
low_limit * 1.3,
dt.strftime('%B'),
rotation=90,
color='firebrick',
alpha=0.6)
class CountyReport:
"""
Generate a county level report showing detailed data on different scenarios.
Parameters
----------
fips: str
County fips code.
model_ensemble: list(SEIRModel)
Models from the ensemble.
county_outputs: dict
Dictionary of structure (in yaml syntax)
suppression_policy__0.5:
t_list: array of timestamps
HICU: For compartment S we have a bunch of properties
S_ci50: the median posterior model. Other condidence intervals are available.
...
peak_value_ciXX:
peak_time_ciXX:
...
# For HICU, HVent, HGen, capacities are relevant, so we also calculate surge windows.
capacity: estimated capacity for this compartment.
surge_start: array surge window starts for each model in the ensemble
surge_end:
filename: str
Where to save the report.
summary: dict
Summary attrs to print on the first page of the report.
xlim: tuple
Limits in days since simulation start to plot.
"""
def __init__(self, fips, model_ensemble, county_outputs, filename, summary, xlim=(0, 360)):
self.fips = fips
self.county_outputs = county_outputs
self.model_ensemble = model_ensemble
self.county_metadata = load_data.load_county_metadata_by_fips(fips)
self.summary = summary
timeseries = combined_datasets.get_timeseries_for_fips(
fips, columns=[CommonFields.CASES, CommonFields.DEATHS], min_range_with_some_value=True
)
self.county_case_data = timeseries.data
self.report = PDFReport(filename=filename)
self.xlim = xlim
def generate_and_save(self):
"""
Name says it all.
"""
self.write_header_pages()
self.plot_seir_distributions()
self.report.close()
def write_header_pages(self):
"""
Generate header pages with the county summary and PDF summaries.
"""
self.report.write_text_page(
self.summary,
title=f'PySEIR COVID19 Estimates\n{self.county_metadata["county"]} County, {self.county_metadata["state"]}',
page_heading=f'Generated {self.summary["date_generated"]}',
body_fontsize=6,
title_fontsize=12,
)
self.report.write_text_page(
inspect.getsource(ParameterEnsembleGenerator.sample_seir_parameters),
title="PySEIR Model Ensemble Parameters",
)
def plot_seir_distributions(self, xlim=(0, 360)):
"""
Generate plots for each suppression policy containing distributions and
peak information for each model compartment.
Parameters
----------
xlim: tuple
Limits in days since simulation start to plot.
"""
# Add a sample model from the ensemble.
fig = self.model_ensemble[0].plot_results(xlim=xlim)
fig.suptitle(
f'PySEIR COVID19 Estimates: {self.county_metadata["county"]} County, {self.county_metadata["state"]}. '
f"SAMPLE OF MODEL ENSEMBLE",
fontsize=16,
)
self.report.add_figure(fig)
suppression_policies = [
key for key in self.county_outputs.keys() if key.startswith("suppression_policy")
]
for suppression_policy in suppression_policies:
output = self.county_outputs[suppression_policy]
compartments = list(output.keys())
compartments.remove("t_list")
# -----------------------------------
# Plot each compartment distribution
# -----------------------------------
fig = plt.figure(figsize=(20, 24))
fig.suptitle(
f'PySEIR COVID19 Estimates: {self.county_metadata["county"]} County, {self.county_metadata["state"]}. '
f"\nSupression Policy={suppression_policy} (1=No Suppression)",
fontsize=16,
)
for i_plot, compartment in enumerate(compartments):
plt.subplot(5, 5, i_plot + 1)
plt.plot(
output["t_list"],
output[compartment]["ci_50"],
color="steelblue",
linewidth=3,
label=compartment_to_name_map[compartment],
)
plt.fill_between(
output["t_list"],
output[compartment]["ci_32"],
output[compartment]["ci_68"],
alpha=0.3,
color="steelblue",
)
plt.fill_between(
output["t_list"],
output[compartment]["ci_5"],
output[compartment]["ci_95"],
alpha=0.3,
color="steelblue",
)
plt.yscale("log")
plt.ylim(1e0)
plt.xlim(0, 360)
plt.grid(True, which="both", alpha=0.3)
plt.xlabel("Days Since Case 0")
# Circular import :(
from pyseir.ensembles.ensemble_runner import compartment_to_capacity_attr_map
if compartment in compartment_to_capacity_attr_map:
percentiles = np.percentile(
[
getattr(m, compartment_to_capacity_attr_map[compartment])
for m in self.model_ensemble
],
(5, 32, 50, 68, 95),
)
plt.hlines(
percentiles[2], *plt.xlim(), label="ICU Capacity", color="darkseagreen"
)
plt.hlines(
[percentiles[0], percentiles[4]],
*plt.xlim(),
color="darkseagreen",
linestyles="-.",
alpha=0.4,
)
plt.hlines(
[percentiles[1], percentiles[3]],
*plt.xlim(),
color="darkseagreen",
linestyles="--",
alpha=0.2,
)
# Plot data
if compartment in ["D", "total_deaths"] and len(self.county_case_data) > 0:
plt.errorbar(
(self.county_case_data.date - self.summary["t0"]).dt.days,
self.county_case_data.deaths,
yerr=np.sqrt(self.county_case_data.deaths),
linestyle="-",
label="Deaths Observed",
marker="o",
markersize=4,
)
if compartment in ["I"] and len(self.county_case_data) > 0:
plt.errorbar(
(self.county_case_data.date - self.summary["t0"]).dt.days,
self.county_case_data.cases,
yerr=np.sqrt(self.county_case_data.cases),
linestyle="-",
label="Cases Observed",
marker="o",
markersize=4,
color="firebrick",
)
plt.legend()
self._plot_dates(log=False)
# -----------------------------
# Plot peak Timing
# -----------------------------
color_cycle = plt.rcParams["axes.prop_cycle"].by_key()["color"] + list("bgrcmyk")
marker_cycle = ["o", "s", "+", "d", "o"] * 4
plt.subplot(5, 5, len(compartments) + 1)
for i, compartment in enumerate(
[
"E",
"A",
"I",
"HGen",
"HICU",
"HVent",
"general_admissions_per_day",
"icu_admissions_per_day",
"direct_deaths_per_day",
"total_deaths_per_day",
]
):
median = output[compartment]["peak_time_ci50"]
ci5, ci95 = (
output[compartment]["peak_time_ci5"],
output[compartment]["peak_time_ci95"],
)
ci32, ci68 = (
output[compartment]["peak_time_ci32"],
output[compartment]["peak_time_ci68"],
)
plt.scatter(
median,
i,
label=compartment_to_name_map[compartment],
c=color_cycle[i],
marker=marker_cycle[i],
)
plt.fill_betweenx(
[i - 0.3, i + 0.3], [ci32, ci32], [ci68, ci68], alpha=0.3, color=color_cycle[i],
)
plt.fill_betweenx(
[i - 0.1, i + 0.1], [ci5, ci5], [ci95, ci95], alpha=0.3, color=color_cycle[i]
)
self._plot_dates(log=False)
plt.legend(loc=(1.05, 0.0))
plt.grid(True, which="both", alpha=0.3)
plt.xlabel("Peak Time After $t_0(C=5)$ [Days]")
plt.yticks([])
# -----------------------------
# Plot peak capacity
# -----------------------------
plt.subplot(5, 5, len(compartments) + 3)
for i, compartment in enumerate(
[
"E",
"A",
"I",
"R",
"D",
"total_deaths",
"direct_deaths_per_day",
"total_deaths_per_day",
"HGen",
"HICU",
"HVent",
"HGen_cumulative",
"HICU_cumulative",
"HVent_cumulative",
"general_admissions_per_day",
"icu_admissions_per_day",
]
):
median = output[compartment]["peak_value_ci50"]
ci5, ci95 = (
output[compartment]["peak_value_ci5"],
output[compartment]["peak_value_ci95"],
)
ci32, ci68 = (
output[compartment]["peak_value_ci32"],
output[compartment]["peak_value_ci68"],
)
plt.scatter(
median,
i,
label=compartment_to_name_map[compartment],
c=color_cycle[i],
marker=marker_cycle[i],
)
plt.fill_betweenx(
[i - 0.3, i + 0.3], [ci32, ci32], [ci68, ci68], alpha=0.3, color=color_cycle[i]
)
plt.fill_betweenx(
[i - 0.1, i + 0.1], [ci5, ci5], [ci95, ci95], alpha=0.3, color=color_cycle[i]
)
plt.xscale("log")
plt.vlines(
self.county_metadata["total_population"],
*plt.ylim(),
label="Entire Population",
alpha=0.5,
color="g",
)
plt.vlines(
self.county_metadata["total_population"] * 0.65,
*plt.ylim(),
label="Approx. Herd Immunity",
alpha=0.5,
color="purple",
linestyles="--",
linewidths=2,
)
plt.legend(loc=(1, -0.1))
plt.grid(True, which="both", alpha=0.3)
plt.xlabel("Value at Peak")
plt.yticks([])
self.report.add_figure(fig)
def _plot_dates(self, log=True):
"""
Helper function to add date plots.
Parameters
----------
log: bool
If True, shift y-positioning of labels based on a log scale.
"""
low_limit = plt.ylim()[0]
if log:
upp_limit = 1 * np.log(plt.ylim()[1])
else:
upp_limit = 1 * plt.ylim()[1]
for month in range(4, 11):
dt = datetime.datetime(day=1, month=month, year=2020)
offset = (dt - self.summary["t0"]).days
plt.vlines(offset, low_limit, upp_limit, color="firebrick", alpha=0.4, linestyles=":")
plt.text(
offset,
low_limit * 1.3,
dt.strftime("%B"),
rotation=90,
color="firebrick",
alpha=0.6,
)