Working with Network Objects in EpiModel

Introduction

This vignette discusses mechanisms usable inside EpiModel network models with custom modules. More information about these in the Extending EpiModel section of the Network Modeling for Epidemics course materials.

Inside the simulation, the networks themselves are stored under dat$run$nw. Ultimately this vignette will describe multiple aspects of working with network objects.

Cumulative Edgelist

The cumulative edgelist refers to the historical list of edges in a network with the time step they start and stopped. Such a list allows to query current relationships (contacts, partnerships, etc.) as well as past ones.

Using the Cumulative Edgelist

The creation and update of the cumulative edgelist is done through the EpiModel::update_cumulative_edgelist function.

dat <- update_cumulative_edgelist(dat, network, truncate = Inf)

This function takes 3 arguments:

dat: the Main List Object.
network: the number of the network for which to create the cumulative edgelist (for multi-layer network simulations).
truncate: a number of time steps after which an inactive edge should be removed from the cumulative edgelist (this saves computer memory for large-scale simulations).

The function returns a modified version of dat that needs to be assigned back.

The following snippet will update the cumulative edgelist for all the networks used by a model and remove the edges that ended more than 100 steps ago.

for (n_network in seq_along(dat$run$nw)) {
  dat <- update_cumulative_edgelist(dat, n_network, truncate = 100)
}

In a complete model, this code would need to be run at the end of the initialization module and at each time-step after the resimulation of the networks.

Accessing the Cumulative Edgelist

Cumulative edge-list refers to nodes with their Unique Ids. See help("unique_id-tools", package = "EpiModel") for more information.

For a Specific Network

Accessing the cumulative edge-list of a given network is done using the EpiModel::get_cumulative_edgelist function.

el_cuml <- get_cumulative_edgelist(dat, network)

The returned el_cuml object is a tibble with four columns:

head: the unique_id first node of the edge.
tail: the unique_id second node of the edge.
start: the time-step where the edge was created.
stop: the last time-step the edge was active.

EpiModel::get_cumulative_edgelist will always return a tibble with this 4 columns, even if the cumulative edgelist has not been calculated for this particular network or if no edges are present. In these cases, the tibble will have no rows but keep the correct column structure.

The stop column will always contain NA if an edge is currently active.

Once an edge is not present anymore, the stop column for this edge will contains the last step the edge was active. This means that an edge with a stop value existed from start to stop both inclusive. This makes it coherent with how R treats the indexes in a vector for instance (from 1 to length(vector) inclusive).

For Multiple Networks

We often want to get the cumulative edgelist over several networks as one. EpiModel::get_cumulative_edgelists_df function provide such functionality.

el_cumls <- get_cumulative_edgelists_df(dat, networks = NULL)

The networks argument can be a vector of network position or NULL. In this latter case, all networks will be selected.

The output of this function is similar to EpiModel::get_cumulative_edgelist with the addition of a network column, indicating for each edge the networks it exists on.

Contact Tracing

A typical use of the cumulative edgelist is the trace the contacts of a node over given number of steps. The EpiModel::get_partners function simplifies this process:

partner_list <- get_partners(
    dat,
    index_posit_ids,
    networks = NULL,
    truncate = Inf,
    only.active.nodes = FALSE
)

Here we call “indexes” the nodes whose partners (contacts) we want to extract. The arguments are:

dat: as in get_cumulative_edgelists_df.
index_posit_ids: a list of positional Ids for the indexes of interest.
networks: as in get_cumulative_edgelists_df.
truncate: similar to the truncate argument to update_cumulative_edgelist this argument filter out partnerships over this age.
only.active.nodes: if set to TRUE, partnership with inactive nodes are removed.

The output is similar to get_cumulative_edgelists_df but the first two columns are called index and partner and contains the Unique Ids of the indexes given in argument in the first column and there partners in the second one.

Note that the we refer to the indexes of interest with their Positional Ids but the index and partners columns contains Unique Ids as they can refer to nodes no longer in the network.

EpiModel v2.5.0