Updates

jls-network-over-time.R

@@ -0,0 +1,64 @@
+library(network)
+library(networkDynamic)
+library(tidyverse)
+library(ndtv)
+
+# Read the data
+data <- read.csv("data/jls-journal-network-openalex.csv") |>
+  filter(citation_year >= 1974)
+
+# Create a lookup table for journal titles to IDs
+titles <- unique(c(data$source_title_citing, data$source_title_cited))
+title_lookup <- tibble(name = titles, id = seq_along(titles))
+
+# Convert source and target titles to ids
+data <- data |>
+  left_join(title_lookup, by = c("source_title_citing" = "name")) |>
+  rename(source_id = id) |>
+  left_join(title_lookup, by = c("source_title_cited" = "name")) |>
+  rename(target_id = id)
+
+# Create the network using an edgelist
+net <- network(matrix(c(data$source_id, data$target_id), ncol = 2), directed = TRUE, loops = FALSE)
+network.vertex.names(net) <- titles
+
+# Create networkDynamic object
+dynNet <- networkDynamic(net)
+
+# Create a unique identifier for each edge in the original data
+data <- data %>% mutate(edge_id_original = paste0(source_id, "-", target_id))
+
+# Initialize a vector to hold the edge IDs from the network object
+edge_ids_network <- numeric(nrow(data))
+
+# Create a progress bar
+pb <- txtProgressBar(min = 0, max = nrow(data), style = 3)
+
+# Loop through the data to find the corresponding edge IDs in the network object
+for (i in seq_len(nrow(data))) {
+  # Update progress bar
+  setTxtProgressBar(pb, i)
+
+  edge <- data[i, ]
+  edge_ids_network[i] <- which.edge(net, tail = edge$source_id, head = edge$target_id)
+}
+
+# Close progress bar
+close(pb)
+
+# Add these edge IDs as a new column to the original data
+data$edge_id_network <- edge_ids_network
+
+# Activate edges
+activate.edges(dynNet, onset = data$citation_year, terminus = data$citation_year, e = data$edge_id_network)
+
+# Add vertex activity
+activate.vertices(dynNet, onset = min(data$citation_year, na.rm = TRUE), terminus = max(data$citation_year, na.rm = TRUE), v = seq_len(network.size(net)))
+
+# Add count_citations as an edge attribute
+set.edge.attribute(dynNet, "count_citations", data$count_citations)
+
+# Render the movie
+render.d3movie(dynNet, launchBrowser = TRUE)
+
+

jls-network.R

@@ -1,4 +1,5 @@
 # see also https://www.r-bloggers.com/2019/06/interactive-network-visualization-with-r/
+# todo: statistical validation of results: https://cran.r-project.org/web/packages/robin/vignettes/robin.html
 
 library(igraph)
 library(tidyr)
@@ -17,25 +18,27 @@ library(dplyr)
 # configuraion
 min_year <- 0
 min_all_years <- 5
-ignore_journals <- c('Sustainability',
-                     'Choice Reviews Online',
-                     'RePEc: Research Papers in Economics',
-                     "DOAJ (DOAJ: Directory of Open Access Journals)")
+louvain_cluster_resolution <- 1
+ignore_journals <- c('sustainability',
+                     'choice reviews online',
+                     'repec: research papers in economics',
+                     "doaj (doaj: directory of open access journals)")
 # data source
 journal_id <- "jls"
-data_vendor <- "openalex"
+data_vendor <- "wos"
 data_file <- paste0("data/", journal_id, "-journal-network-", data_vendor, ".csv")
 
 # dataframe with columns source_title1, source_title2, count_citations
-df <- read.csv(data_file)
-
-# remove self-citations
-df <- df |>
-  filter(!(source_title_citing %in% ignore_journals) & !(source_title_cited %in% ignore_journals)) |>
+df <- read.csv(data_file) |>
+  # remove journals that have a high citation rate but are not relevant for our question
+  filter(!(str_to_lower(source_title_citing) %in% ignore_journals) &
+           !(str_to_lower(source_title_cited) %in% ignore_journals)) |>
+  # remove self-citations
   filter(source_title_citing != source_title_cited) |>
+  # remove journals which do not meet a minimum of citations per year
   filter(count_citations >= min_year)
 
-# Calculate total citations made by each journal per year and normalized weight
+# Calculate total citations made by each journal per year and compute normalized weight
 citing_totals_per_year <- df |>
   group_by(source_title_citing, citation_year) |>
   summarise(total_citing = sum(count_citations))
@@ -45,8 +48,7 @@ df <- df |>
 
 # Create list of nodes with id and label
 all_nodes <- data.frame(label = unique(c(df$source_title_citing, df$source_title_cited))) |>
-  mutate(id=seq_along(label)) |>
-  select(id, label)
+  mutate(id=seq_along(label))
 
 # Create list of edges
 edges <- df |>
@@ -74,16 +76,17 @@ edges <- edges |>
     from = first(from),
     to = first(to),
     total_count = sum(count),
-    combined_median_weight = median(c(rep(median_weight, count))),
+    combined_weight = median(weight, na.rm = TRUE),
     label_AB = sum(if(from < to) count else 0),  # A->B
     label_BA = sum(if(from > to) count else 0)   # B->A
   ) |>
   # Create the desired label and weight
   mutate(
     label = paste(label_AB, "/", label_BA),
-    weight = combined_median_weight
+    weight = combined_weight,
+    size = combined_weight * 10
   ) |>
-  select(from, to, label, weight)
+  select(from, to, label, weight, size)
 
 # Filter out nodes without edges
 edges_node_ids <- unique(c(edges$from, edges$to))
@@ -99,7 +102,7 @@ largest_comp_id <- which.max(comps$csize)
 graph <- induced_subgraph(graph, which(comps$membership == largest_comp_id))
 
 # Louvain Comunity Detection
-cluster <- cluster_louvain(graph)
+cluster <- cluster_louvain(graph, resolution = louvain_cluster_resolution)
 cluster_groups <- membership(cluster)
 cluster_df <- tibble(group=cluster_groups) |>
   mutate(id = as.integer(row_number())) |>