`community.prism()`

Pairwise Resolution Iteration via Subgraph Modularity (PRISM)

Description

Iteratively tests pairwise unions of communities for a modularity gain at the subgraph level and merges communities whenever the union increases modularity relative to the communities' separate memberships. Iteration continues until no pair of communities yields a consistent, positive gain, at which point the partition is returned. The approach operates on single subgraphs rather than the full network at each comparison, which is intended to overcome the modularity resolution limit described by Fortunato and Barthelemy (2007)

Usage

community.prism(
  network,
  algorithm = c("edge_betweenness", "fast_greedy", "fluid", "infomap", "label_prop",
    "leading_eigen", "leiden", "louvain", "optimal", "spinglass", "walktrap"),
  allow.singleton = FALSE,
  seed = NULL,
  ...
)

Arguments

`network`	Matrix or `igraph` network object
`algorithm`	Character or `igraph` `cluster_` function (length = 1). Available options: `"edge_betweenness"` — See `cluster_edge_betweenness` for more details `"fast_greedy"` — See `cluster_fast_greedy` for more details `"fluid"` — See `cluster_fluid_communities` for more details `"infomap"` — See `cluster_infomap` for more details `"label_prop"` — See `cluster_label_prop` for more details `"leading_eigen"` — See `cluster_leading_eigen` for more details `"leiden"` — See `cluster_leiden` for more details. Note*: The Leiden algorithm will default to the modularity objective function (`objective_function = "modularity"`). Set `objective_function = "CPM"` to use the Constant Potts Model instead (see examples) `"louvain"` — Applies `EGAnet`'s own C implementation of the Louvain algorithm (Blondel et al., 2008), which replaces `cluster_louvain`. Accepts `resolution` (defaults to `1`), `seed` (defaults to `NULL`, i.e., not reproducible), and `order` (`"higher"` or `"lower"`; defaults to `"higher"`). Unlike `igraph::cluster_louvain`, which draws on R's global random number generator, this implementation's randomness depends only on `seed` – making it reproducible regardless of R's RNG state, call order, or parallelization `"optimal"` — See `cluster_optimal` for more details `"spinglass"` — See `cluster_spinglass` for more details `"walktrap"` — See `cluster_walktrap` for more details
`allow.singleton`	Boolean (length = 1). Whether singleton or single node communities should be allowed. Passed on to the function selected via `algorithm`. When `FALSE`, singleton communities will be set to missing (`NA`); otherwise, when `TRUE`, singleton communities will be allowed
`seed`	Numeric (length = 1). Sets seed for reproducible results. Defaults to `NULL` or random results. Passed on to the function selected via `algorithm` (relevant for `algorithm = "louvain"`) for every pairwise subgraph comparison performed during the iterative procedure, making the entire PRISM procedure reproducible
`...`	Additional arguments to be passed on to the community detection function selected via `algorithm` (see `community.consensus` or `community.detection` for their respective arguments)

Value

Returns a named vector of final community memberships obtained by iterating pairwise resolution over subgraph modularity until convergence (i.e., until no further consistent modularity gain is found)

Author(s)

Alexander P. Christensen <alexpaulchristensen@gmail.com>

References

Granell, C., Gomez, S., & Arenas, A. (2012). Hierarchical multiresolution method to overcome the resolution limit in complex networks. International Journal of Bifurcation and Chaos, 22(07), 1250171.

Fortunato, S., & Barthelemy, M. (2007). Resolution limit in community detection. Proceedings of the National Academy of Sciences, 104(1), 36-41.

Examples

# Load data
wmt <- wmt2[,7:24]

# Estimate network
network <- EBICglasso.qgraph(data = wmt)

# Compute PRISM memberships (Louvain-based consensus)
community.prism(network)

# Compute PRISM memberships (single-algorithm community detection)
community.prism(network, algorithm = "walktrap")