Simulate a trajectory from a SimInf model. The
function compiles and loads model-specific C code (if not already
compiled), initializes the solver, and advances the simulation in
continuous time from the first to the last time point in
tspan. The state of the system is recorded at each time
point specified in tspan. Sparse output can be requested
with punchcard<-
to store only selected time points or compartments.
Usage
run(model, ...)
# S4 method for class 'SimInf_model'
run(model, ...)
# S4 method for class 'SEIR'
run(model, ...)
# S4 method for class 'SIR'
run(model, ...)
# S4 method for class 'SIS'
run(model, ...)
# S4 method for class 'SISe'
run(model, ...)
# S4 method for class 'SISe3'
run(model, ...)
# S4 method for class 'SISe3_sp'
run(model, ...)
# S4 method for class 'SISe_sp'
run(model, ...)
# S4 method for class 'SimInf_abc'
run(model, ...)Arguments
- model
The SimInf model to run.
- ...
Optional arguments that affect the simulation:
- solver
Character string specifying the numerical solver. Either
"ssm"(default) or"aem". If not provided, the"ssm"solver is used. See details.- seed
Numeric or integer specifying the random seed for the simulation. If not provided, a seed is randomly sampled from the current R RNG state.
- crn
Logical. If
TRUE, use common random numbers for the"mssm"solver by assigning each node and replicate a separate random number stream, enabling variance reduction across simulation runs. If not provided,crnisFALSE.- rng_type
Character string specifying the random number generator algorithm. Either
"mt19937"or"taus2". If not provided,"mt19937"is used.
Details
The solver uses a split-step method: for each unit of time, it first integrates the continuous-time Markov chain within each node using direct SSA (Gillespie's algorithm), then processes scheduled events (exit, enter, internal transfer, and external transfer), and finally calls the post time step function to update continuous state variables; see Widgren and others (2019) for details.
Two numerical solvers are available. The default solver is
"ssm" (split-step method), which becomes "mssm"
(multi-model split-step method) automatically when the model
contains multiple replicates (replicates > 1). The
alternative solver is "aem" (all events method), which
assigns each reaction channel its own random number stream for
consistent operational times across simulations; see Bauer and
Engblom (2015). The "aem" solver cannot be used with
replicated models.
References
S. Widgren, P. Bauer, R. Eriksson and S. Engblom. SimInf: An R Package for Data-Driven Stochastic Disease Spread Simulations. Journal of Statistical Software, 91(12), 1–42, 2019. doi:10.18637/jss.v091.i12 . An updated version of this paper is available as a vignette in the package.
P. Bauer, S. Engblom and S. Widgren. Fast Event-Based Epidemiological Simulations on National Scales. International Journal of High Performance Computing Applications, 30(4), 438–453, 2016. doi: 10.1177/1094342016635723
P. Bauer and S. Engblom. Sensitivity Estimation and Inverse Problems in Spatial Stochastic Models of Chemical Kinetics. In: A. Abdulle, S. Deparis, D. Kressner, F. Nobile and M. Picasso (eds.), Numerical Mathematics and Advanced Applications - ENUMATH 2013, pp. 519–527, Lecture Notes in Computational Science and Engineering, vol 103. Springer, Cham, 2015. doi:10.1007/978-3-319-10705-9_51
Examples
## For reproducibility, specify the number of threads to use.
set_num_threads(1)
## Create an 'SIR' model with 10 nodes.
model <- SIR(
u0 = data.frame(
S = rep(99, 10),
I = rep(1, 10),
R = rep(0, 10)
),
tspan = 1:100,
beta = 0.16,
gamma = 0.077
)
## Run the model with a fixed seed for reproducibility.
result <- run(model, seed = 22)
## Plot the distribution of susceptible, infected and recovered
## individuals.
plot(result)
