analyze lots of networks

analyze.all.networks.r

library(data.table)
library(igraph)

x <- fread("all_networks_140923.csv")

igraph.options(vertex.label = NA, edge.arrow.size = 0, vertex.size = 5)

g <- graph.data.frame(x[,list(source, target, weight, type, survey)])

### Graph level statistics ################################################### size
# edges
# density
# components
# apl 
# modularity


### filter the graph
graph.stats <- function(edge.type, edge.survey) {
  g1 <- subgraph.edges(g, which(E(g)$type == edge.type & E(g)$survey == edge.survey), delete.vertices = T)
  g1 <- simplify(as.undirected(g1))
  m <- edge.betweenness.community(g1)
  bet <- betweenness(g1)
  data.table(type = edge.type, survey = edge.survey,
             size = vcount(g1), 
             edge.count = length(E(g1)),
             density = graph.density(g1),
             components = clusters(g1)$no,
             avg.path.length = average.path.length(g1),
             modularity = modularity(m),
             centralization = centralization.degree(g1)$centralization,
             assortativity = assortativity.degree(g1),
             bet.m = mean(bet, na.rm=T)
             bet.sd = sd(bet, na.rm=T))
}
survey.u <- x[,.N, by=list(survey,type)][,N := NULL]

graph.stats("know", "fu13")

all.stats <- data.table(type = character(), survey = character(),
                        size = numeric(), edge.count = numeric(),
                        density = numeric(), components = numeric(),
                        avg.path.length = numeric(), modularity = numeric(),
                        centralization = numeric(),
                        assortativity = numeric())


for(i in 1:6) {
  cat("Running stats for ", survey.u$type[i], " and ", survey.u$survey[i], "\n", sep="")
  all.stats <- rbind(all.stats, 
                     graph.stats(edge.type = survey.u$type[i],
                                 edge.survey = survey.u$survey[i]))
}

#cat("Running stats for zipcode ", i, " of ", length(zip.u), "\n", sep="")

xx <- do.call('rbind', mclapply(1:nrow(survey.u), 
       function(i) { 
         graph.stats(edge.type = survey.u$type[i], edge.survey = survey.u$survey[i])
       }, mc.cores = 6))


gs <- list()
for(i in 1:nrow(survey.u)) {
  gs[[i]] <- subgraph.edges(g, which(E(g)$type == survey.u$type[i] & E(g)$survey == survey.u$survey[i]), delete.vertices = T)
}

lapply(gs, "graph.density")



# izip = indexed zip
izip <- 40517
g <- induced.subgraph(a, vids = which(V(a)$zip3 == izip))



### Node level stats ##########################################################
# centralities