#Network analysis of ICD-11 CPTSD symptoms

#Load data 
library(haven)
data <-read_sav("data")
View(data)

#Loading other libraries for later use
library("qgraph")
library("bootnet") 
library("corpcor")
library("ggplot2")

#Rename variables and for this analysis CPTSD symptoms 
names(data)[names(data) == "itq2_p1"] <- "RE1"
names(data)[names(data) == "itq2_p2"] <- "RE2"
names(data)[names(data) == "itq2_p3"] <- "AV1"
names(data)[names(data) == "itq2_p4"] <- "AV2"
names(data)[names(data) == "itq2_p5"] <- "TH1"
names(data)[names(data) == "itq2_p6"] <- "TH2"
names(data)[names(data) == "itq2_c1"] <- "AD1"
names(data)[names(data) == "itq2_c2"] <- "AD2"
names(data)[names(data) == "itq2_c3"] <- "NSC1"
names(data)[names(data) == "itq2_c4"] <- "NSC2"
names(data)[names(data) == "itq2_c5"] <- "DR1"
names(data)[names(data) == "itq2_c6"] <- "DR2"
names(data)

#Labels ITQ
legend <- c("Distressing dreams","Intrusive recollections/flashbacks", 
            "Internal avoidance", "External avoidance",
            "Hypervigilance/Sense of current threat",
          "Exaggerated startle response","Long-time upset (hyperactivation)",
          "Emotional numbing (hypoactivation)",
          "Feelings of failure", "Feelings of worthlessness", 
          "Feeling distant or cut off from others", 
          "Difficulities feeling close to others")

#Constructing a new dataset with only ITQ symptoms 
Data1 <- data[c("RE1", "RE2", "AV1", "AV2", "TH1",
                "TH2", "AD1", "AD2", "NSC1", "NSC2", "DR1", "DR2")]

#1 Sample Descriptives (See Table 1 in article)
library(summarytools)
view(dfSummary(data))

#vector - Employment_Status
Employment_Status <- vector("character", length = nrow(data))

#Descriptives for Employment status 
for (i in 1:nrow(data)) {
  if (data$q0005_0013[i] == 1) {
    Employment_Status[i] <- "On Sick Leave"
  } else if (data$q0005_0007[i] == 1 | data$q0005_0008[i] == 1) {
    Employment_Status[i] <- "Student"
  } else if (data$q0005_0009[i] == 1 | data$q0005_0011[i] == 1) {
    Employment_Status[i] <- "Retired"
  } else if (data$q0005_0001[i] == 1 | data$q0005_0003[i] == 1 | data$q0005_0004[i] == 1 | data$q0005_0005[i] == 1 | data$q0005_0006[i] == 1) {
    Employment_Status[i] <- "Employed"
  } else if (data$q0005_0002[i] == 1 | data$q0005_0012[i] == 1 | data$q0005_0016[i] == 1 | data$q0005_0010[i] == 1 | data$q0005_0014[i] == 1 | data$q0005_0015[i] == 1) {
    Employment_Status[i] <- "Other"
  } else {
    Employment_Status[i] <- "Unknown"
  }
}

#Overview - Employment status
view(dfSummary(Employment_Status))

#2 Estimation of the regularized partial correlation network of CPTSD symptoms
library(qgraph)
ggm <- estimateNetwork(Data1, default = "EBICglasso", verbose = FALSE)

#2.1 Extracting the adjacency matrix
adjacency_matrix_GGM <- getWmat(ggm)
View(adjacency_matrix_GGM)

#2.1.1 The same matrix as extracted above, however now with 3 decimals
View(round(adjacency_matrix_GGM, 3))

#2.2 Defining my groups
groups_list <- list(
  RE = c("RE1", "RE2"),
  AV = c("AV1", "AV2"),
  TH = c("TH1", "TH2"),
  AD = c("AD1", "AD2"),
  NSC = c("NSC1", "NSC2"),
  DR = c("DR1", "DR2")
)

#Colorblind-friendly for the groups
node_colors <- c(
  "#E69F00", "#E69F00",  # RE1, RE2
  "#56B4E9", "#56B4E9",  # AV1, AV2
  "#009E73", "#009E73",  # TH1, TH2
  "#F0E442", "#F0E442",  # AD1, AD2
  "#0072B2", "#0072B2",  # NSC1, NSC2
  "#D55E00", "#D55E00"   # DR1, DR2
)
  
#2.3 Plotting the regularized ggm
qgraph(adjacency_matrix_GGM)

#2.4# Making the ggm more fancy by using a colorblind-friendly theme and 
#the layout spring
qgraph(adjacency_matrix_GGM, 
       layout = "spring",
       color = node_colors,
       legend = FALSE,
       legend.cex = 0.4,
       label.cex = 1.0,
       vsize = 4,
       border.width = 2.0, 
       theme = "colorblind",
       GLratio = 1.4,
       legend.cex = .35)

#Saving my ggm plot as a pfd
pdf("Regularized GGM CPTSD_2.pdf")
qgraph(adjacency_matrix_GGM, 
       layout = "spring",
       color = node_colors,
       legend = FALSE,
       legend.cex = 0.4,
       label.cex = 1.0,
       vsize = 4,
       border.width = 2.0, 
       theme = "colorblind",
       GLratio = 1.4,
       legend.cex = .35)

#3 Network description - centrality strenght

#Standardized centrality strenght
Strenght <- centralityPlot(adjency_matrix_GGM, scale = "z-scores")

#Saving the strenght plot as pdf
pdf("Standardized_strenght.pdf")
centralityPlot(adjency_matrix_GGM, scale = "z-scores")
dev.off()

#####Stability analysis for edge weight and centrality strenght########

#4.1 Checking accuracy of the estimated edge weights
#by using **nonparametric bootstrapping with replacement** to construct 95% bootstrapped confidence intervals.

#Setting seed for reproducibility
set.seed(1)

#Running the non-parametric bootstrap
boot.edges <- bootnet(ggm, boots = 1000, nCore = 8, verbose = FALSE)

#Plotting the 95% bootstrapped confidence intervals
plot(boot.edges, labels = TRUE, order = "sample")

#Saving the plot as a pfd
pdf("BOOTNET1.pdf")
plot(boot.edges, labels = TRUE, order = "sample")
dev.off()

#Edge weights difference plot
plot(boot.edges, "edge", plot = "difference", onlyNonZero = TRUE, 
     order = "sample")

#Saving the plot as a pdf
pdf("BOOTNET2_DIFFERENCETEST.pdf")
plot(boot.edges,  "edge", plot = "difference", onlyNonZero = TRUE, order = "sample")
dev.off()

#4.2 Checking centrality stability
#by using the **case-dropping subset bootstrap** 

#Setting seed for reproducibility
set.seed(2)

#Running the case-dropping subset bootstrap 
boot.centrality <- bootnet(ggm, boots = 1000, nCores = 8, type = "case", verbose = FALSE)

#Centrality bootstrap plot
plot(boot.centrality) 

#Save the plot as a pdf
pdf("BOOTNET3CENTRALITY.pdf")
plot(boot.centrality)
dev.off()

#Centrality difference plot for strength centrality 
plot(boot.edges, "strength", plot = "difference")

#Save the plot as a pdf
pdf("BOOTNET4CENTRALITYDIFFERENCETEST.pdf")
plot(boot.edges, "strength", plot = "difference")
dev.off()

#Numerical results of the CS-coefficient
Correlation_koefficienter <- corStability(boot.centrality)
round(Correlation_koefficienter, 2)
View(Correlation_koefficienter)

#Done