R/sample_ic.R

In princyparsana/netsimulatR:

## construct a sparse inverse covariance matrix
sample_ic <- function(p = 10, m=1, psd_type = "jgl", u = 0.1, v = 0.3, type = "scale-free" ,min_support = c(-0.4, 0.1), max_support = c(-0.1, 0.4), seed = 1){
  set.seed(seed)
  if(type == "scale-free"){
    pg = igraph::sample_pa(p, power = 2.5, m = m, directed = F)
  }else{
    pg = igraph::sample_gnm(p, m = p*3)
  }
  n_edges = igraph::ecount(pg)
  g <- igraph::as_adj(pg)
  g <- Matrix::triu(g)
  e_vals = runif(n_edges, min = min_support, max = max_support)
  g[g!=0] <- e_vals
  g <- convert_psd(g, u = u, v = v, psd_type = psd_type)
  if(!all(eigen(g)$values >=0)){
    cat("PSD check unsuccessful - negative eigenvalues")
  }else{
    cat("PSD check was successful - all eigenvalues are greater than or equal to 0")
  }
  list(graph = pg, theta = g)
}