Get data

(1) Web of science data

Get dois (Obtained from: https://apps.crossref.org/SimpleTextQuery/ (copy-pasting 50 at a time))

dois <- read.csv("dois2.csv") %>%
      select(study_ID, doi) %>%
      filter(doi != "" & !is.na(doi)) %>%
      mutate(doi = ifelse(grepl(".org/", doi),
                          unlist(lapply(str_split(doi, ".org/"), 
                                   function(x) {x[2]})), as.character(doi)))

Scrape data from web of science at paper level using bibliometrix package

string = ""
for (i in 1:length(dois$doi)) {
  string = paste(string, dois$doi[i], sep = " OR ")
}

## search web of science-> save to marked list -> download marked list as .txt
D <- readFiles("savedrecs (3).txt")
M <- convert2df(D, dbsource = "isi", format = "plaintext") %>%
  mutate(DI = tolower(DI),
         DI = str_replace(DI, "//", "/")) %>%
  arrange(DI) %>%
  filter(!is.na(DI)) %>%
  mutate_each(funs(as.factor), -AU, -DE, -ID)
## Articles extracted   100 
## Articles extracted   163
M <- metaTagExtraction(M, Field = "AU_CO", sep = ";")

Get ES and ns at paper level

paper.data = all_data %>%
  inner_join(dois, by="study_ID") %>%
  mutate(doi = tolower(doi)) %>%
  group_by(doi) %>%
  summarize(d_calc = mean(d_calc, na.rm = T),
            short_name = short_name[1], 
            n = sum(n_1, na.rm = T) + sum(n_2, na.rm = T)) 

# get web of science data for domain
ns = paper.data %>%
  inner_join(M, by = c("doi" = "DI")) %>%
  group_by(short_name) %>%
  summarize(n.papers = n()) 

kable(ns)
short_name n.papers
catBias 7
gaze_following 12
idspref 13
inphondb-native 23
inphondb-nonnative 4
inworddb 49
labadv 13
mutex 16
phonotactics 2
pointing_concurrent 4
pointing_longitudinal 1
sounds 6
symbolism 5
word_recognition 6 
paper.data = paper.data %>%
  left_join(ns) %>%
  filter(n.papers > 2) %>%
  filter(short_name != "pointing_longitudinal")

n.mas = length(unique(paper.data$short_name))

(2) network plot data

getGraphData <- function (my.paper.data, 
                        this.analysis,
                        this.network,
                        remove.multiple = FALSE, 
                        remove.isolates = FALSE,
                        noloops = FALSE) {
    
    # FOR DEBUGGING
      #my.paper.data = filter(paper.data, short_name == unique(paper.data$short_name)[2])
    # this.analysis = "co-occurrences"
    # this.network = "keywords"
  
    # get web of science data for domain
    my.M = left_join(my.paper.data, M, by = c("doi" = "DI")) %>%
             as.data.frame() # necessary for biblioNetwork
    
    SEP = ifelse(this.analysis == "co-citation", ". ", "; ")
  
    # make the network
    this.net <- biblioNetwork(my.M, 
                   analysis = this.analysis, 
                   network = this.network, 
                   sep = SEP) 
    
    if (length(this.net > 0)) {
      # munge the network
      bsk.network <- graph.adjacency(this.net, mode = "undirected")
      V(bsk.network)$id <- colnames(this.net)
      deg <- igraph::degree(bsk.network, mode = "all")
      V(bsk.network)$size = rep(5, length(V(bsk.network)))
  
      n = dim(this.net)[1]
      NetDegree <- unname(sort(deg, decreasing = TRUE)[n])
      bsk.network <- igraph::delete.vertices(bsk.network,
                                     which(igraph::degree(bsk.network) < 
          NetDegree))
      bsk.network <- igraph::simplify(bsk.network, 
                              remove.multiple = remove.multiple, 
          remove.loops = noloops)
      if (isTRUE(remove.isolates)) {
          bsk.network <- igraph::delete.isolates(bsk.network, mode = "in")
      }
      
      # make into df for plotting
      gn = asNetwork(bsk.network)
    }
    
    cc <- try(ggnetwork(gn), silent = TRUE)  # gets rid of too-small networks
    if(!is(cc,"try-error")) {
       return(ggnetwork(gn))
    } 
}

(3) network measures data

getModularity <- function (my.paper.data, 
                        this.analysis,
                        this.network,
                        remove.multiple = FALSE, 
                        remove.isolates = FALSE,
                        noloops = FALSE) {

    # get web of science data for domain
    my.M = left_join(my.paper.data, M, by = c("doi" = "DI")) %>%
             as.data.frame() # necessary for biblioNetwork
    
    SEP = ifelse(this.analysis == "co-citation", ". ", "; ")
  
    # make the network
    this.net <- biblioNetwork(my.M, 
                   analysis = this.analysis, 
                   network = this.network, 
                   sep = SEP) 
    
    if (length(this.net > 0)) {
      # munge the network
      graph <- graph.adjacency(this.net, mode = "undirected")
      clustering = cluster_leading_eigen(graph, 
                              options = list(maxiter=1000000))
    
      closeness = mean(estimate_closeness(graph, cutoff= 100))
      betweeness = mean(estimate_betweenness(graph, cutoff= 100))
      degree = mean(igraph::degree(graph))

      data.frame(analysis = this.analysis,
                network = this.network,
                Q = round(clustering$modularity,4), 
                n.groups = round(length(clustering),4),
                closeness  = round(closeness, 4),
                betweeness  = round(betweeness, 4),
                degree = mean(degree))
    }
}

# get network measures for all networks
analyses <- rep(c("co-citation", "coupling", "co-occurrences", "collaboration"), each = n.mas) 
networks <- rep(c("references", "authors", "keywords", "authors"), each = n.mas)

dfs <- paper.data %>% 
  mutate_each(funs(as.factor)) %>%
  split(.$short_name) %>%
  rep(4)

args = list(dfs, analyses, networks)

modularity.data = args %>%
  pmap(getModularity) %>%
  bind_rows(.id = "short_name")

modularity.data.long = modularity.data %>%
    mutate(closeness = log(closeness),
            betweeness = log(betweeness),
            degree = log(degree)) %>%
    gather("network.measure", "network.value", 4:8) %>%
    unite(network, analysis, network, sep = ".")
  
ggplot(modularity.data.long, aes(x = short_name, y = network.value)) +
        geom_bar(stat = "identity", aes(fill = short_name))  +
        facet_wrap(network.measure ~ network, 
                   ncol = 4, scales = "free") +
        theme_bw() +
        theme(axis.text.x = element_blank())

(4) literature bias data

ma.es = read.csv("ES_data_for_networks.csv") %>%
 # mutate(d_var = ci_upper-ci_lower)%>%
  select(short_name, overall.d, overall.d.age,
         fsn_string, egg.random.z, stouffer.Z.pp) %>%
  gather("bias.measure", "bias.value", 2:6) 
  #filter(bias.measure %in% c("egg.random.z", "fsn_string"))

Networks and correlations with bias

Get model fits

d = left_join(modularity.data.long, ma.es) %>%
  dplyr::as_data_frame() %>%
  left_join(ns)

corrs = d %>%
  filter(is.finite(network.value) & is.finite(bias.value)) %>%
  group_by(network, bias.measure, network.measure) %>%
  #do(tidy(cor.test(.$bias.value, .$value, na.action = "na.pass"))) %>%
  do(tidy(lm(bias.value ~ n.papers + network.value, data=.))) %>%
  filter(term == "network.value") %>%
  mutate(sig.col = ifelse(p.value < .05, "sig", "nsig")) %>%
  select(estimate, sig.col) %>%
  mutate(network.value = Inf, bias.value = Inf)  %>% # this is a hack
  as.data.frame()

Collaboration authors

Networks

ANALYSIS <-  "collaboration"
paper.data %>%
  split(.$short_name) %>%
  map(getGraphData, ANALYSIS , "authors") %>%
  bind_rows(.id = "short_name") %>%
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
     geom_edges(color = "grey50") +
     facet_wrap(~short_name) +
     geom_nodes(aes(color = short_name)) +
     ggtitle("authors") +
     theme_blank() +
     theme(legend.position = "none")

Correlations

ggplot(filter(d, network == "collaboration.authors"), 
       aes(x = network.value, y = bias.value)) +
  geom_rect(data = filter(corrs, network == "collaboration.authors"),
            aes(fill = sig.col), 
          xmin = -Inf, xmax = Inf,
          ymin = -Inf, ymax = Inf, alpha = 0.2) +
  geom_point(aes(size = n.papers, color = short_name)) +
  geom_smooth(method= "lm") +
  ggtitle("collaboration.authors")+
  facet_grid(bias.measure ~ network.measure, scales = "free") +
  scale_fill_manual(values = c( "grey99", "red1")) +
  theme_bw() +
  theme(legend.position = "none")

Co-citation references

Networks

ANALYSIS <-  "co-citation"
paper.data %>%
  split(.$short_name) %>%
  map(getGraphData, ANALYSIS, "references") %>%
  bind_rows(.id = "short_name") %>%
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
     geom_edges(color = "grey50") +
     facet_wrap(~short_name) +
     geom_nodes(aes(color = short_name)) +
     ggtitle("references") +
     theme_blank() +
     theme(legend.position = "none")

Correlations

ggplot(filter(d, network == "co-citation.references"), 
       aes(x = network.value, y = bias.value)) +
  geom_point(aes(size = n.papers, color = short_name)) +
  geom_smooth(method= "lm") +
  ggtitle("co-citation.references") +
  facet_grid(bias.measure ~ network.measure, scales = "free") +
  geom_rect(data = filter(corrs, network == "co-citation.references"),
            aes(fill = sig.col), 
          xmin = -Inf, xmax = Inf,
          ymin = -Inf, ymax = Inf, alpha = 0.2) +
  scale_fill_manual(values = c( "grey99", "red1")) +
  theme_bw() +
  theme(legend.position = "none")

## Coupling references{.tabset}
### Networks
ANALYSIS <-  "coupling"
paper.data %>%
  split(.$short_name) %>%
  map(getGraphData, ANALYSIS, "references") %>%
  bind_rows(.id = "short_name") %>%
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
     geom_edges(color = "grey50") +
     facet_wrap(~short_name) +
     geom_nodes(aes(color = short_name)) +
     ggtitle("references") +
     theme_blank() +
     theme(legend.position = "none")

Coupling authors

Networks

ANALYSIS <-  "coupling"
paper.data %>%
  split(.$short_name) %>%
  map(getGraphData, ANALYSIS, "authors") %>%
  bind_rows(.id = "short_name") %>%
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
     geom_edges(color = "grey50") +
     facet_wrap(~short_name) +
     geom_nodes(aes(color = short_name)) +
     ggtitle("authors") +
     theme_blank() +
     theme(legend.position = "none")

Correlations

ggplot(filter(d, network == "coupling.authors"), 
       aes(x = network.value, y = bias.value)) +
  geom_rect(data = filter(corrs, network == "coupling.authors"),
            aes(fill = sig.col), 
          xmin = -Inf, xmax = Inf,
          ymin = -Inf, ymax = Inf, alpha = 0.2) +
  geom_point(aes(size = n.papers, color = short_name)) +
  geom_smooth(method= "lm") +
  ggtitle("coupling.authors") +
  facet_grid(bias.measure ~ network.measure, scales = "free") +

  scale_fill_manual(values = c( "grey99", "red1")) +
  theme_bw() +
  theme(legend.position = "none")

Co-occurences keywords

Networks

ANALYSIS <-  "co-occurrences"
paper.data %>%
  split(.$short_name) %>%
  map(getGraphData, ANALYSIS, "keywords") %>%
  bind_rows(.id = "short_name") %>%
  ggplot(aes(x = x, y = y, xend = xend, yend = yend)) +
     geom_edges(color = "grey50") +
     facet_wrap(~short_name) +
     ggtitle("keywords") +
     geom_nodes(aes(color = short_name)) +
     theme_blank() +
     theme(legend.position = "none")

Correlations

ggplot(filter(d, network == "co-occurrences.keywords"), 
       aes(x = network.value, y = bias.value)) +
  geom_point(aes(size = n.papers, color = short_name)) +
  geom_smooth(method= "lm") +
  ggtitle("co-occurrences.keywords") +
  facet_grid(bias.measure ~ network.measure, scales = "free") +
  geom_rect(data = filter(corrs, network == "co-occurrences.keywords"),
            aes(fill = sig.col), 
          xmin = -Inf, xmax = Inf,
          ymin = -Inf, ymax = Inf, alpha = 0.2) +
  scale_fill_manual(values = c( "grey99", "red1")) +
  theme_bw() +
  theme(legend.position = "none")

Paper level analyses

getModularity_paper <- function (my.paper.data, 
                        this.analysis,
                        this.network,
                        remove.multiple = FALSE, 
                        remove.isolates = FALSE,
                        noloops = FALSE) {

    # get web of science data for domain
    my.M = left_join(my.paper.data, M, by = c("doi" = "DI")) %>%
             as.data.frame() # necessary for biblioNetwork
    
    SEP = ifelse(this.analysis == "co-citation", ". ", "; ")
  
    # make the network
    this.net <- biblioNetwork(my.M, 
                   analysis = this.analysis, 
                   network = this.network, 
                   sep = SEP) 
    
    if (length(this.net > 0)) {
      # munge the network
      graph <- graph.adjacency(this.net, mode = "undirected")
      data.frame(analysis = this.analysis,
                 network = this.network,
                 closeness = estimate_closeness(graph, cutoff= 100),
                 betweeness = estimate_betweenness(graph, cutoff= 100),
                 degree = igraph::degree(graph))
    }
}

dfs <- paper.data %>% 
  mutate_each(funs(as.factor)) %>%
  split(.$short_name) %>%
  rep(4)
analyses <- rep(c("co-citation", "coupling", "co-occurrences", "collaboration"), each = 14) 
networks <- rep(c("references", "authors", "keywords", "authors"), each = 14)
args = list(dfs, analyses, networks)

modularity.data = args %>%
  pmap(getModularity_paper) %>%
  filter(analysis == "co-citation") %>%
  bind_rows(.id = "short_name")
k = M %>%
    left_join(paper.data, by=c("DI"= "doi")) %>%
    split(.$short_name) %>%
    map(biblioAnalysis) %>%
    bind_rows() 

bibliometrix::summary(k, pause = FALSE) 


S = summary(object = m, k = 10, pause = FALSE)

TO DO:

  • get summary statistics using summary function
  • get ES controling for age, and ES error, also domain bias, n
  • correlate everything at domain and entire dataset
  • look at individual papers