Plotting¶

The plotting functions live in epydemix.visualization. They are built on matplotlib: each one accepts an optional ax to draw on, and returns the axes so you can keep customising the figure.

from epydemix.visualization import plot_quantiles, plot_trajectories

Function	Shows
`plot_quantiles()`	median and interval over time, optionally with observed data
`plot_trajectories()`	every individual run
`plot_posterior_distribution()`	posterior of one parameter
`plot_posterior_distribution_2d()`	joint posterior of two parameters
`plot_distance_distribution()`	distances of the accepted simulations
`plot_population()`	population by demographic group
`plot_contact_matrix()`	contact matrix of one layer
`plot_spectral_radius()`	intensity of contacts over time, with interventions

Simulation results¶

from epydemix import load_predefined_model

model = load_predefined_model("SIR")
results = model.run_simulations(start_date="2024-01-01", end_date="2024-04-10", Nsim=100)

# Median and 90% interval
plot_quantiles(
    results.get_quantiles_compartments(),
    columns=["Infected_total", "Recovered_total"],
    labels=["Infected", "Recovered"],
    ylabel="People",
)

# Every run as a thin line
plot_trajectories(
    results.get_stacked_compartments(),
    columns="Infected_total",
    dates=results.dates,
)

The interval drawn by plot_quantiles is set by lower_q and upper_q (5% and 95% by default), which must be among the quantiles in the table you pass. To overlay observed data, pass a DataFrame as data with a date column and a column named like the variable you plot, and set show_data=True, as in Calibration.

Several plots in one figure¶

import matplotlib.pyplot as plt

fig, axes = plt.subplots(1, 2, figsize=(11, 4))
plot_quantiles(results.get_quantiles_compartments(), columns="Infected_total", ax=axes[0], title="Infected")
plot_quantiles(results.get_quantiles_transitions(), columns="Susceptible_to_Infected_total", ax=axes[1], title="New infections")

Other plots¶

Populations and contact matrices: see Populations and contacts.
Contact intensity over time: see Interventions.
Posterior distributions and model fit: see Calibration.

For plots the built-in functions don’t cover, get the data out as a table with results.to_dataframe() (see Running simulations) and use matplotlib or seaborn directly.