netsampler {EcoNetGen}

Network Sampling Routine

Description

Network Sampling Routine

Usage

netsampler(network_in, key_nodes_sampler = c("random", "lognormal",
  "Fisher log series", "exponential", "degree", "module"),
  neighbors_sampler = c("random", "exponential"), n_key_nodes = 10,
  n_neighbors = 0.5, hidden_modules = NULL, module_sizes = NULL,
  cluster_fn = igraph::cluster_edge_betweenness)

Arguments

network_in

input network (as igraph object)

key_nodes_sampler

sampling criteria for key nodes. See details.

neighbors_sampler

sampling criteria for neighbors. see details.

n_key_nodes

number of key nodes to sample.

n_neighbors

number of first neighbors or fraction of first neighbors. See details.

hidden_modules

list of the modules to exclude (max 10 modules; only the first numb_hidden are used)

module_sizes

integer vector giving the size of each module. see details.

cluster_fn

a clustering function, from igraph::cluster_*. Default is igraph::cluster_edge_betweeness. Only used to compute module sizes if not provided.

Details

Algorithm first samples n_key_nodes according the the requested key_nodes_sampler criterion. For each key node, the requested number or fraction of neighbors is then sampled according to the neighbors_sampler criterion. Optionally, a list of modules can be designated as "hidden" and will be excluded from sampling.

if n_neighbors is greater than 1, assumes this is the number to sample. If n_neighborsis between 0 and 1, assumes this is the fration of neighbors to sample. (To sample 1 neighbor, use an explicit integer,1L (or as.integer(1)') to sample 100

Provide module_sizes list to improve performance. If not provided, this will will be calculated based on igraph::cluster_edge_betweeness. Be sure to provide a module_sizes vector whenever calling netsampler repeatedly on the same network to avoid unnecessary performance hit from recalculating modules every time. See examples.

Value

the original input network (as an igraph network object), with the attribute label added to the edges and vertices indicating if that edge or vertex was sampled or unsampled.

Examples

set.seed(12345)
net <- netgen()
sample <- netsampler(net)

## Precompute `module_sizes` for replicate sampling of the same network:
 library(igraph)
 modules <- cluster_edge_betweenness(as.undirected(net))
 module_sizes <- vapply(igraph::groups(modules), length, integer(1))
 sample <- netsampler(net, module_sizes = module_sizes)