DarkNet-Syn.md

Step 1. Install and load required packages for correlations and partial correlations graphs.

---

install.packages("qgraph") #see qgraph: Network Visualizations of Relationships in Psychometric Data for details

library(qgraph) #load qgraph

Step 2. Load simulated HDS data and store as correlation matrix (pearson or tetrachoric).

---

For illustration purposes, code is illustrated using the pearson correlation matrix.

library(psych)
HDS.Sim.cor <- cor(SimData.HDS) #Pearson
HDS.Sim.tet <- tetrachoric(SimData.HDS) #Tetrachoric

Step 3. Plot correlation and partial correlation networks (figures 2A and 2B).

---

groups <- c(rep("Narcissism", 14), rep("Psychoticism", 14), rep("Machievallianism", 14)) #color nodes by scale

##Correlation network##
netCor <- qgraph(HDS.Sim.cor, layout = "spring", labels = colnames(HDS.Sim.cor), groups = groups, 
vsize=4, title = "A. Correlation Network", legend = F)

##Partial correlation network - Cut edge correlations < .10##
netPcor <- qgraph(HDS.Sim.cor, layout = "spring", labels = colnames(HDS.Sim.cor), groups = groups, graph = "concentration", 
vsize = 4, title = "B. Partial Correlation Network", legend = F, cut = .10, esize = 10)

Step 4. Plot Adaptive LASSO network - preferred when modeling multiple associations (figure 2c).

---

install.packages("parcor") #Package for regularized estimation of partial correlation matrices

library(parcor)

Estimate the adaptive lasso network¹.

set.seed(100)
adls <- adalasso.net(HDS.Sim) #compute partial correlation matrix via LASSO technique
network <- as.matrix(forceSymmetric(adls$pcor.adalasso)) #extract adaptive LASSO matrix
colnames(network) <- rownames(network) <- colnames(HDS.Sim)

Plot adaptive lasso network

netal <- qgraph(network, layout = "spring", labels = colnames(HDS.Sim), groups=groups, vsize=4, Title = "C. Adaptive LASSO Network", legend = F, cut = .10, esize = 10)

Create a uniform layout based upon the adapative lasso network (see Figure 2)

L <- netal$layout
qgraph(netCor, layout = L)
qgraph(netPcor, layout = L)

Step 5. Extracting, describing, and testing differences in LASSO partial correlations (Table 2).

---

Creating partial correlation table (Table 2)

ew <- round(network,2)
ew[upper.tri(ew)] <- ""
ew <- as.data.frame(ew)

Descriptives and t-test for comparing difference in weights between positive v. negative edges

ew <- network[upper.tri(network)]
describe(ew)
sum(ew != 0)
sum (ew > 0)
sum (ew < 0)
t.test(abs (ew [ew > 0]), abs(ew [ew < 0]), var.equal = TRUE)

Step 6. Calculating and visualizing global and local network properties (Figures 3-4).

---

smallworldness(network) #Calculate small-worldness index (Humphries & Gurney, 2008)

centrality <- centrality_auto(network) #Calculate and store centrality indices (Betweenness, closeness, strength)

centralityPlot(network, theme_bw = FALSE) #Visualize centrality indices

cluster <- clustcoef_auto(network) #Calculate and store clustering coefficients

clusteringPlot(network, theme_bw = FALSE) #Visualize multiple clustering coefficients

Notes

¹ qgraph implements a similiar yet slightly distinct EBICglasso function.