R: Control Settings for Stochastic Network Models

control.net {EpiModel}

R Documentation

Control Settings for Stochastic Network Models

Description

Sets the controls for stochastic network models simulated with netsim.

Usage

control.net(
  type,
  nsteps,
  start = 1,
  nsims = 1,
  ncores = 1,
  resimulate.network = FALSE,
  tergmLite = FALSE,
  cumulative.edgelist = FALSE,
  truncate.el.cuml = 0,
  attr.rules,
  epi.by,
  initialize.FUN = initialize.net,
  resim_nets.FUN = resim_nets,
  summary_nets.FUN = summary_nets,
  infection.FUN = NULL,
  recovery.FUN = NULL,
  departures.FUN = NULL,
  arrivals.FUN = NULL,
  nwupdate.FUN = nwupdate.net,
  prevalence.FUN = prevalence.net,
  verbose.FUN = verbose.net,
  module.order = NULL,
  save.nwstats = TRUE,
  nwstats.formula = "formation",
  save.transmat = TRUE,
  save.network,
  save.other,
  verbose = TRUE,
  verbose.int = 1,
  skip.check = FALSE,
  raw.output = FALSE,
  tergmLite.track.duration = FALSE,
  set.control.ergm = control.simulate.formula(MCMC.burnin = 2e+05),
  set.control.tergm = control.simulate.formula.tergm(),
  save.diss.stats = TRUE,
  dat.updates = NULL,
  ...
)

Arguments

`type`	Disease type to be modeled, with the choice of `"SI"` for Susceptible-Infected diseases, `"SIR"` for Susceptible-Infected-Recovered diseases, and `"SIS"` for Susceptible-Infected-Susceptible diseases.
`nsteps`	Number of time steps to simulate the model over. This must be a positive integer that is equal to the final step of a simulation. If a simulation is restarted with `start` argument, this number must be at least one greater than that argument's value.
`start`	For models with network resimulation, time point to start up the simulation. For restarted simulations, this must be one greater than the final time step in the prior simulation and must be less than the value in `nsteps`.
`nsims`	The total number of disease simulations.
`ncores`	Number of processor cores to run multiple simulations on, using the `foreach` and `doParallel` implementations.
`resimulate.network`	If `TRUE`, resimulate the network at each time step. This is required when the epidemic or demographic processes impact the network structure (e.g., vital dynamics).
`tergmLite`	Logical indicating usage of either `tergm` (`tergmLite = FALSE`), or `tergmLite` (`tergmLite = TRUE`). Default of `FALSE`.
`cumulative.edgelist`	If `TRUE`, calculates a cumulative edgelist within the network simulation module. This is used when tergmLite is used and the entire networkDynamic object is not used.
`truncate.el.cuml`	Number of time steps of the cumulative edgelist to retain. See help for `update_cumulative_edgelist` for options.
`attr.rules`	A list containing the rules for setting the attributes of incoming nodes, with one list element per attribute to be set (see details below).
`epi.by`	A character vector of length 1 containing a nodal attribute for which subgroup stratified prevalence summary statistics are calculated. This nodal attribute must be contained in the network model formation formula, otherwise it is ignored.
`initialize.FUN`	Module to initialize the model at time 1, with the default function of `initialize.net`.
`resim_nets.FUN`	Module to resimulate the network at each time step, with the default function of `resim_nets`.
`summary_nets.FUN`	Module to extract summary statistics of the network at each time step, with the default function of `summary_nets`.
`infection.FUN`	Module to simulate disease infection, with the default function of `infection.net`.
`recovery.FUN`	Module to simulate disease recovery, with the default function of `recovery.net`.
`departures.FUN`	Module to simulate departure or exit, with the default function of `departures.net`.
`arrivals.FUN`	Module to simulate arrivals or entries, with the default function of `arrivals.net`.
`nwupdate.FUN`	Module to handle updating of network structure and nodal attributes due to exogenous epidemic model processes, with the default function of `nwupdate.net`.
`prevalence.FUN`	Module to calculate disease prevalence at each time step, with the default function of `prevalence.net`.
`verbose.FUN`	Module to print simulation progress to screen, with the default function of `verbose.net`.
`module.order`	A character vector of module names that lists modules in the order in which they should be evaluated within each time step. If `NULL`, the modules will be evaluated as follows: first any new modules supplied through `...` in the order in which they are listed, then the built-in modules in the order in which they are listed as arguments above. `initialize.FUN` will always be run first and `verbose.FUN` will always be run last.
`save.nwstats`	If `TRUE`, save network statistics in a data frame. The statistics to be saved are specified in the `nwstats.formula` argument.
`nwstats.formula`	A right-hand sided ERGM formula that includes network statistics of interest, with the default to the formation formula terms. Supports `multilayer` specification.
`save.transmat`	If `TRUE`, complete transmission matrix is saved at simulation end.
`save.network`	If `TRUE`, networkDynamic or networkLite object is saved at simulation end.
`save.other`	A character vector of elements on the `netsim_dat` main data list to save out after each simulation. One example for base models is the attribute list, `"attr"`, at the final time step.
`verbose`	If `TRUE`, print model progress to the console.
`verbose.int`	Time step interval for printing progress to console, where `0` prints completion status of entire simulation and positive integer `x` prints progress after every `x` time steps. The default is to print progress after each time step.
`skip.check`	If `TRUE`, skips the default error checking for the structure and consistency of the parameter values, initial conditions, and control settings before running base epidemic models. Setting this to `FALSE` is recommended when running models with new modules specified.
`raw.output`	If `TRUE`, `netsim` will output a list of raw data (one per simulation) instead of a cleaned and formatted `netsim` object.
`tergmLite.track.duration`	If `TRUE`, track duration information for models in `tergmLite` simulations. Supports `multilayer` specification.
`set.control.ergm`	Control arguments passed to `ergm::simulate_formula.network`. In `netsim`, this is only used when initializing the network with `edapprox = TRUE`. All other simulations in `netsim` use `tergm`. Supports `multilayer` specification.
`set.control.tergm`	Control arguments passed to `tergm::simulate_formula.network`. See the help file for `netdx` for details and examples on specifying this parameter. Supports `multilayer` specification.
`save.diss.stats`	If `TRUE`, `netsim` will compute and save duration and dissolution statistics for plotting and printing, provided `save.network` is `TRUE`, `tergmLite` is `FALSE`, and the dissolution model is homogeneous.
`dat.updates`	Either `NULL`, a single function taking arguments `dat`, `at`, and `network`, or a list of functions of length one greater than the number of networks being simulated, with each function in the list taking arguments `dat` and `at`. Here `dat` is the main `netsim_dat` class object, `at` is the current timestep, and `network` is an index indicating the current position within the sequence of network (re)simulations on each time step. If a single function is passed, it will be called before the first network is simulated and after each network is simulated, with `network = 0L` before the first network is simulated and with `network = i` after the `i`th network is simulated. If a list of functions is passed, the first function will be called before the first network is simulated, and the `i + 1`th function will be called after the `i`th network is simulated. (Note that `at = 0L` is used for initial cross-sectional simulations in `sim_nets_t1`.) The function(s) should return the `netsim_dat` object with any updates needed to correctly represent the network states for calls to `simulate` and/or `summary`. This can be useful if nodal attributes appearing in one network model depend on nodal degrees in a different network.
`...`	Additional control settings passed to model.

Details

control.net sets the required control settings for any network model solved with the netsim function. Controls are required for both base model types and when passing original process modules. For an overview of control settings for base models, consult the Basic Network Models tutorials. For all base models, the type argument is a necessary parameter and it has no default.

Value

An EpiModel object of class control.net.

The attr.rules Argument

The attr.rules parameter is used to specify the rules for how nodal attribute values for incoming nodes should be set. These rules are only necessary for models in which there are incoming nodes (i.e., arrivals). There are three rules available for each attribute value:

current: new nodes will be assigned this attribute in proportion to the distribution of that attribute among existing nodes at that current time step.
t1: new nodes will be assigned this attribute in proportion to the distribution of that attribute among nodes at time 1 (that is, the proportions set in the original network for netest).
Value: all new nodes will be assigned this specific value, with no variation. For example, the rules list attr.rules = list(race = "t1", sex = "current", status = "s") specifies how the race, sex, and status attributes should be set for incoming nodes. By default, the rule is "current" for all attributes except status, in which case it is "s" (that is, all incoming nodes are susceptible).

Checkpointing Simulations

netsim has a built-in checkpoint system to prevent losing computation work if the function is interrupted (SIGINT, power loss, time limit exceeded on a computation cluster). When enabled, each simulation will be saved every .checkpoint.steps time steps. Then, if a checkpoint enabled simulation is launched again with netsim, it will restart at the last checkpoint available in the saved data.

To enable the checkpoint capabilities of netsim, two control arguments have to be set: .checkpoint.steps, which is a positive number of time steps to be run between each file save; and .checkpoint.dir, which is the path to a directory to save the checkpointed data. If .checkpoint.dir directory does not exist, netsim will attempt to create it on the first checkpoint save. With these two controls defined, one can simply re-run netsim with the same arguments to restart a set of simulations that were interrupted.

Simulations are checkpointed individually: for example, if 3 simulations are run on a single core, the first 2 are finished, then the interruption occurs during the third, netsim will only restart the third one from the last checkpoint.

A .checkpoint.compress argument can be set to overwrite the compress argument in saveRDS used to save the checkpointed data. The current default for saveRDS is gunzip (gz), which provides fast compression that usually works well on netsim objects.

By default, if netsim reaches the end of all simulations, the checkpoint data directory and its content are removed before returning the netsim object. The .checkpoint.keep argument can be set to TRUE to prevent this removal to inspect the raw simulation objects.

New Modules

Base network models use a set of module functions that specify how the individual nodes in the network are subjected to infection, recovery, demographics, and other processes. Core modules are those listed in the .FUN arguments. For each module, there is a default function used in the simulation. The default infection module, for example, is contained in the infection.net function.

For original models, one may substitute replacement module functions for any of the default functions. New modules may be added to the workflow at each time step by passing a module function via the ... argument. Consult the New Network Models tutorials. One may remove existing modules, such as arrivals.FUN, from the workflow by setting the parameter value for that argument to NULL.

End Horizon

netsim implements an "End Horizon" mechanism, where a set of modules are removed from the simulation at a specific time step. This is enabled through the end.horizon parameter to control.net.

This parameter must receive a list with fields at, the time step at which the end horizon occurs, and modules, a character vector with the names of the modules to remove. (e.g 'list(at = 208, modules = c("arrivals.FUN", "infections.FUN")))