Given a raw co-occurrence matrix (e.g. from compute_cooc_matrix()),
computes expected frequencies under independence and retains only
edges where observed / expected > 1. The diagonal and upper triangle
are zeroed so each pair appears at most once.
Arguments
- cooc_matrix
A symmetric numeric matrix (dense or sparse) of co-occurrence counts, as returned by
compute_cooc_matrix().
Value
A data.table with columns from, to, weight containing
only edges whose observed co-occurrence exceeds the independence
baseline. Edges are sorted by descending weight.
Examples
if (FALSE) { # \dontrun{
mat <- compute_cooc_matrix(skills)
rr_edges <- filter_relative_risk(mat)
} # }