Skip to contents

Keep the main component(s) of networks

Source: R/filter_components.R
filter_components.Rd

[Experimental]

A projected network is often made of one large connected component and of small islands: a few articles that share references only with each other, or isolated pairs. Community detection and layouts are hard to read on such fragments, and many studies keep only the main component. filter_components() numbers the (weakly) connected components by decreasing size, keeps the nb_components largest, and warns when the first component removed is not negligible.

Usage

filter_components(
  graphs,
  nb_components = 1,
  threshold_alert = 0.05,
  keep_component_columns = FALSE
)

Arguments

graphs

A tbl_graph or a list of tbl_graphs (e.g. the output of build_dynamic_networks()).

nb_components

Integer, the number of components to keep, starting from the largest (1 by default).

threshold_alert

Number between 0 and 1. If the largest removed component gathers more than this share of the nodes (0.05 by default), a warning is given: removing it would change the object of study.

keep_component_columns

Logical. If TRUE, keeps the columns components_att (component rank, 1 being the largest) and size_components (share of the nodes in the component).

Value

The same object as graphs (a tbl_graph or a list of tbl_graphs) restricted to the nodes of the kept components.

Details

Dropping small components is a choice, not a technicality: the removed documents may be a marginal literature, or a literature that cites sources absent from the data. Look at what is removed before discarding it, especially when the warning is triggered. For a list of graphs, the filter is applied to each graph separately, so a node may be kept in one window and removed in another.

See also

build_network() and build_dynamic_networks(), which can call this function with filter_components = TRUE.

Examples

network <- build_network(
  nodes = subset(Nodes_stagflation, source_type == "Stagflation"),
  directed_edges = Ref_stagflation,
  source_id = "source_id",
  target_id = "target_id",
  method = "similarity",
  edges_threshold = 2
)
#> ℹ Coupling (`biblionetwork::biblio_coupling()`) with "similarity": ties
#>   weighted by the idf-weighted cosine of the two lists of targets, after Shen
#>   et al. (2019), between source_id entities sharing target_id entities.
#> ℹ Two source_id are linked when they share at least 2 target_id
#>   (`edges_threshold`).
#> ℹ Nodes without any edge are removed (`keep_singleton = FALSE`).
main_component <- filter_components(network, nb_components = 1)
c(before = igraph::vcount(network), after = igraph::vcount(main_component))
#> before  after 
#>    131    129