Events

Motivation

Events allow to model infections which have exogenous causes and occur besides contacts to infectious people. Examples of such events are

• Imported cases which caused the center of infections in Munich when ski-tourists returned.

• A carnival session which sparked the outbreak in Gangelt.

• The RKI models a steady inflow of one infection per day.

Specifying an event

Multiple events are stored in a dictionary. Each event has a name in this dictionary like "gangelt". The values of the dictionary are again dictionaries which specify the event and could have the following keys.

{
    "gangelt": {
        # "loc": slice(None),  # Returns all parameters which is the default.
        "model": carnival_session,
    }
}

"loc"

Expression to select a subset of params. This is mostly relevant if pre-implemented events are used (e.g. import_cases) and the params can be used to select covariates from states. The same contact model could be used with a different parameterization. This key is optional.

"model"

A model is a function which receives states, params, and a seed and returns a boolean series where new infections are marked with True. The new infections per day from contacts and multiple events are merged with a logical OR.

import numpy as np
import pandas as pd

from sid import get_date


def carnival_session(states, params, seed):
    np.random.seed(seed)
    date = get_date(states)  # Helper to get the current date from states.
    date_of_carnival_session = pd.Timestamp("2020-02-15")

    newly_infected = pd.Series(data=False, index=states.index)

    if date == date_of_carnival_session:

        adults_in_heinsberg = states.query(
            "county == 'Kreis Heinsberg' and age_group >= '10 - 19'"
        ).index

        infection_prob = params.loc[("infection_prob", "gangelt", None), "value"]
        infected_indices = np.random.choice(
            adults_in_heinsberg, size=300 * infection_prob, replace=False
        )

        newly_infected.loc[infected_indices] = True

    return newly_infected