01_edges.r

library(dplyr)

edges = bind_rows(
  data_frame(i = "A", j = c("A", "B", "C"), w = 2),
  data_frame(i = "B", j = c("B", "A"), w = 1),
  data_frame(i = "A", j = c("A", "C"), w = 1),
  data_frame(i = "B", j = c("B", "C", "D", "E"), w = 3)
)

02_self_loops.r

# first author self-loops, with counts of co-authors
self = filter(edges, i == j)

# count number of texts per first author
n_au = table(self$i)

03_coauthorship.r

# remove self-loops from directed edge list
edges = filter(edges, i != j)

# count number of texts coauthored per co-author
n_co = table(edges$j)

04_raw_counts.r

# collapse directed ties
edges$ij = apply(edges[, 1:2 ], 1, paste0, collapse = "->")

# raw edge counts
raw = table(edges$ij)

05_newman_fowler_weights.r

# Newman-Fowler weights
edges = aggregate(w ~ ij, function(x) sum(1 / x), data = edges)

06_re_expand.r

# re-expand to edge list
edges = data_frame(i = gsub("(.*)->(.*)", "\\1", edges$ij),
                   j = gsub("(.*)->(.*)", "\\2", edges$ij),
                   raw = as.vector(raw[ edges$ij ]), # raw edge counts
                   nfw = edges$w)

07_gross_kirkland_shalizi_weights.r

# Gross-Kirkland-Shalizi weights
edges = left_join(edges, aggregate(w ~ i, function(x) sum(1 / x), data = self))
edges$gsw = edges$nfw / edges$w

08_sanity_check.r

# sanity check
stopifnot(edges$gsw >= 0 & edges$gsw <= 1)

09_finalize.r

# final edge list: first author, co-author, edge weights
edges = select(edges, -w)