R/defn.R

In njetzel/ngc: Methods for Estimating Graphical Granger Causality

#' Creates the skeleton of a random sparse network
#' @param d lag of VAR process
#' @param p dimension of VAR process
#' @param n sample size
#' @param weight weights of edges to be sampled
#' @param grp group structure of covariates
#' @return constructed network
#' @export defn_net
defn_net =
  function(
    d,
    p,
    n,
    weight = c(1, 1, 1),
    grp = NULL
  ){
    if (is.null(grp))
    {
      grp <- 1:p
    }
    edge = array(0, c(p, p, d))
    signum = c(1, -1)
    grpCt = length(unique(grp))
    sparsity = min((n/(d*grpCt*p)), 0.05)
    cat(paste("sparsity =", round(sparsity, 4)))
    for (ii in 1:d){
      for (i in 1:p){
        for (j in unique(grp)){
            edge[i, grp == j, ii] = ((runif(1, 0, 1) < sparsity))*sample(weight, 1)*sample(signum, 1)
        }
      }
    }
    return(edge)
  }