sim_trajectory_CSV_decoupled {MGDrivE2}

Simulate Trajectory From a SPN Model

Description

This function provides a unified interface to the various simulation algorithms for SPN, returning output sampled at a lattice of time points to the user, and handling various exogenous events that may occur during the simulation (such as release of adult mosquitoes). This function is used in decoupled sampling, where the mosquito and human states are separated. This is used primarily when using the Imperial model of malaria transmission.

Usage

sim_trajectory_CSV_decoupled(
  x0,
  h0,
  inf_labels,
  tmax,
  dt = 1,
  dt_stoch = 0.1,
  folders = "./",
  S,
  hazards,
  SPN_P,
  theta,
  Sout = NULL,
  sampler = "tau",
  method = "lsoda",
  events = NULL,
  batch = NULL,
  verbose = TRUE,
  human_ode = "Imperial",
  cube = cube,
  ...
)

Arguments

x0

the initial marking of the SPN (initial state, M0)

h0

the initial human state distribution

inf_labels

labels corresponding to female mosquito infection hazard

tmax

the final time to end simulation

dt

the time-step at which to return output (not the time-step of the sampling algorithm)

dt_stoch

time-step used for approximation of hazards

folders

vector of folders to write output

S

a stoichiometry Matrix-class object

hazards

list of hazard functions

SPN_P

stochastic petri net places

theta

parameters

Sout

an optional matrix to track event firings

sampler

determines sampling algorithm, one of; "ode", "tau", "cle", or "dm"

method

if sampler is "ode", the solver to use, from deSolve

events

a data.frame of events

batch

a list of batch migration events, created from batch_migration, may be set to NULL if not used

verbose

print a progress bar?

human_ode

human ode function

cube

inheritance cube

...

further named arguments passed to the step function

Details

dt_stoch is used by the Poisson Time-Step (step_PTS) and Chemical Langevin (step_CLE) methods to approximate the hazards. A smaller dt_stoch provides a better approximation, but will take longer to run.

The stoichiometry matrix (S) is generated in spn_S.

The list of hazards (hazards) come from spn_hazards.

Several samplers are provided. The default is a Poisson Time-Step (step_PTS) method. Other options are Gillespie's Direct Method (step_DM) and a Chemical Langevin sampler (step_CLE). Additionally, for convenience, an ODE "sampler" (step_ODE) is provided for compatibility with other samplers. This function uses methods from deSolve.

If using the ode sampler, several methods are provided in the deSolve package, see ode. For inhomogeneous systems, consider using the "rk4" method to avoid excessive integration times.

Additionally, events objects must follow the format required by deSolve. This was done for consistency, see events for more information.

This function writes all output to .csv files. Each simulation is written to a folder element - the number of repetitions is the number of folders provided. For now, only adult mosquito states, human states, clinical incidence, and pathogen prevalence are written to CSVs.

This function tracks state variables specified by argument stage by default; an optional argument Sout can be provided to track number of event firings each time step (for discrete stochastic simulations), or cumulative intensity (for continuous stochastic simulations), or the rate function of particular events for ODE simulation. The matrix must have number of columns equal to number of events in the system (the number of hazard functions), and a row for each tracking variable. If Sout is provided, it outputs an additional csv, "Tracking.csv". The function track_hinf is provided, which builds a matrix to track human infection events.

To return simulations to R for further processing, see sim_trajectory_R.

Value

NULL - prints output to .csv files