Exact simulation of a compartmental model and its surveillance counts

Gillespie simulation of the Markov jump process defined by model, with counts aggregated over intervals of length Delta and reported counts drawn as binomial thinning plus Gaussian measurement error (rounded and truncated at zero), the data-generating process assumed by ga_loglik().

Usage

ga_simulate(model, theta, init_counts, obs, n, Delta = 1, varying = NULL)

Arguments

model

A ga_model().

theta

Named list or vector with the rate parameters of model and any parameters named in obs.

init_counts

Named integer vector of initial counts per compartment.

obs

List of ga_incidence() / ga_prevalence() streams.

n

Number of intervals.

Delta

Length of the observation interval.

varying

Optional named list of rate parameters that change between intervals, each a vector of length n (as in ga_loglik()); a rate listed here may be absent from theta.

Value

A list with data (matrix of reported counts, one column per stream), latent (matrix of latent counts per stream) and states (compartment counts at the end of each interval, first row the initial state).