Generate a random graph with a given structure — generate

Draws a random undirected graph on p nodes from one of the available structure generators, rejecting degenerate draws (the empty graph and the complete graph) and resampling until a usable structure is obtained. The returned adjacency matrix is what simulate_prior_study uses for simulating prior study parameters.

Usage

generate_graph(
  p,
  structure = c("smallworld", "random", "scalefree", "Bernoulli"),
  ...,
  max_attempts = 1000L
)

Arguments

p: Number of nodes (variables) in the graph. Must be at least 3.
structure: Which graph generator to use: "smallworld", "random", "scalefree" or "Bernoulli".
...: Further arguments passed to the graph generator. The "Bernoulli" structure requires prob, the edge-inclusion probability in \((0, 1)\). The other structures take no extra arguments.
max_attempts: Maximum number of rejection-sampling draws before giving up with an error. Guards against parameter choices for which a non-degenerate graph is effectively unreachable.

Value

A symmetric p by p adjacency matrix, 0/1 entries, zero diagonal.

Details

Generation uses rejection sampling: a draw is accepted only if it has at least one edge and is not fully connected. Since no non-degenerate graph exists for p < 3, p must be at least 3

Examples

set.seed(2026)
# Bernoulli (Erdos-Renyi) needs an edge probability
g <- generate_graph(p = 10, structure = "Bernoulli", prob = 0.2)
dim(g)
#> [1] 10 10

# \donttest{
g2 <- generate_graph(p = 10, structure = "smallworld")
# }