4 Importing and Visualizing One- and Two-Mode Social Network Data in statnet
In this lab we’ll explore a variety of methods for importing social network data into R, manipulating one- and two-mode network data, and visualizing social networks. We’ll be using a variety of social networks, some of which you’ll recognize from other classes. We’ll also illustrate a variety of ways to import network data, something that should be easy to do but often turns out to be challenging because a number of resources jump over this important step.
Note: This lab has gone through many iterations and reflects the influence from a variety of individuals, including Phil Murphy, and Brendan Knapp.
4.1 Setup
Find and open your RStudio Project associated with this class. Begin by opening a new script. It’s generally a good idea to place a header at the top of your scripts that tell you what the script does, its name, etc.
#######################################################################
# What: Importing and Visualizing One- and Two-Mode Social Network Data
# File: lab1_statnet.R
# Created: 02.28.14
# Revised: 01.05.22
#######################################################################If you have not set up your RStudio Project to clear the workspace on exit, your environment contain the objects and functions from your prior session. To clear these before beginning use the following command.
rm(list = ls())Proceed to place the data required for this lab (davis.csv, davis.net, davisedge.csv, Koschade Bali (Edge).csv, Koschade Bali (Matrix).csv, and Koschade Bali.net) also inside your R Project folder. We have placed it in a sub folder titled data for organizational purposes; however, this is not necessary.
4.2 Load Libraries
We need to load the libraries we plan to use. Here we will use statnet. Because igraph and statnet conflict with one another sometimes, we do not want to have them loaded at the same time, so you may want to detach it. Alternatively, you may choose to namespace functions using the :: operator as needed (e.g., igraph::betweenness() vs. sna::betweenness()). Of course, this applies only if you had the igraph package loaded already. The intergraph package allows users to transform network data back and forth between igraph and statnet.
# If you haven't done so, install the required packages:
# install.packages("statnet")
# install.packages("intergraph")
# Now load them:
library(statnet)
library(intergraph)4.3 One-mode Social Network Data in statnet: Koschade Network
Here, we will use data collected by Stuart Koschade of the 17 individuals who participated in the first Bali bombing. Koschade (2006) recorded both the ties between the individuals, as well as the strength of the tie between them.
4.3.2 Plotting (Visualizing) the Koschade Network
Plotting in statnet is fairly straight forward. The primary function is gplot(), which produces a two dimensional network visualization and allows you to control vertex placements, edge characteristics, colors, etc. We suggest that you take a quick look at the documentation using the command ?gplot.
To get us started with let’s compare a base visualization against a much more refined graph. The first uses statnet’s defaults, for the second we will modify many of the arguments. In particular, the second tells R that the network is a one-mode network (gmode = graph rather than digraph, which is the default), adds labels using the network.vertex.names() function , colors the labels black, places them in the center of the nodes (label.pos = 5), and changes their size (label.cex = 1.6). The next series of arguments set the size and color of the vertices, hides the arrows, and colors the ties (edges) gray. Note that we saved the coordinates so both plots would have the same layout
# Set graph parameters to 1 row and 2 columns
par(mfrow = c(1, 2))
# Save coordinates in an object
coords <- network.layout.kamadakawai(koschade1_net,
# The function expects a list of parameters
# pass a NULL to use defaults
layout.par = NULL)
# Plot base graph
gplot(koschade1_net,
coord = coords)
# Plot graph using vertex coordinates and additional arugments
gplot(koschade1_net,
gmode = "graph",
coord = coords,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
In the previous visualizations, we used the Kamada and Kawai algorithm to layout the nodes. By default, gplot() uses the Fruchterman and Reingold algorithm to determine the positions of nodes. Let’s compare the visual output of three layout algorithms: Kamada and Kawai, Fruchterman and Reigold, and circle. Please note that many other layouts exist, for a more indepth list look at the documenation ?gplot.layout.
# Set graph parameters to 1 row and 3 columns
par(mfrow = c(1, 3))
# Kamada and Kawai
gplot(koschade1_net,
gmode = "graph",
mode = "kamadakawai",
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
# Kamada and Kawai
gplot(koschade1_net,
gmode = "graph",
mode = "fruchtermanreingold",
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
# Circle
gplot(koschade1_net,
gmode = "graph",
mode = "circle",
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
Before we move forward, let’s take a look at three more arguments that can grately improve the look of your graphs. First, the jitter argument insures that gplot() does not draw vertices on top of one another. Second, remember that the edge and vertex attributes can be called and used to aid the visuals. Here we use the get.edge.attribute() function to call the edge weight vector (1) and rescale the thickness of these. Finally, we can curve edges setting usecurve to TRUE.
gplot(koschade1_net,
gmode = "graph",
coord = coords,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray",
# New arguments
jitter = TRUE,
edge.lwd = get.edge.attribute(koschade1_net, "1"),
usecurve = TRUE,
edge.curve = .1)
4.3.3 Saving Network Plots (e.g., pdf, jpeg, png, tiff)
Save final plot in various formats. Begin by saving the output in PDF format. To do such, use the pdf() function, which starts the graphics driver for producing PDFs.
# Start the graphic driver, name output file, and set size
pdf(file = "koschade1.pdf",
width = 4, height = 4)
# Plot the output into the file
gplot(koschade1_net,
gmode = "graph",
coord = coords,
jitter = TRUE,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
# Turn off the graphics driver
dev.off()To store the image as a JPEG, use the jpeg() function. The bg = "transparent" option saves the graphs with a transparent background (rather than white), which can be helpful when placing in slides or on non-white backgrounds.
jpeg(file = "koschade1.jpg",
width = 4, height = 4,
units = 'in',
res = 600,
bg = "transparent")
gplot(koschade1_net,
gmode = "graph",
coord = coords,
jitter = TRUE,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
dev.off()To store the image as a PNG, use the png() function.
png(file = "koschade1.png",
width = 4, height = 4,
units = 'in',
res = 300,
bg = "transparent")
gplot(koschade1_net,
gmode = "graph",
coord = coords,
jitter = TRUE,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
dev.off()To store the image as a TIFF, use the tiff() function.
tiff(file = "koschade3.tif",
width = 4, height = 4,
units = 'in',
res = 300,
bg = "transparent")
gplot(koschade1_net,
gmode = "graph",
coord = coords,
jitter = TRUE,
label = network.vertex.names(koschade1_net),
label.col = "black",
label.pos = 5,
label.cex = 0.5,
vertex.cex = 1.6,
vertex.col = "light blue",
usearrows = FALSE,
edge.col = "gray")
dev.off()4.3.4 Saving Network Data
Finally, it doesn’t hurt to save the data that you’ve imported and created. Perhaps not all (e.g., coordinates) but it is helpful to save those that you may want to use in another setting.
save(koschade_el,
koschade1_net,
koschade2_net,
koschade3_net,
file = "koschade_statnet.RData")4.4 Two-Mode Social Network Data in statnet: Davis Southern Women
We will now switch to another data set to import, manipulate, and visualize two-mode network data in statnet. The data that we will use here is what is known as Davis’ Southern Club Women. Davis and her colleagues recorded the observed attendance of 18 Southern women at 14 different social events.
Recall that in two-mode network ties only exist between modes. That means that ties are only possible between women and events, not between women and women or between events and events. Any direct ties between nodes within a mode may be derived (projected), as we will do below. But they should not appear within the original network.
4.4.1 Importing Two-Mode Network Data
4.4.1.1 Option 1: Importing Two-Mode Network Data in Matrix Format
Let’s begin my importing two-mode network data that’s recorded in matrix format (i.e., an incidence matrix).
davis_mat <- as.matrix(
read.csv("data/davis.csv",
header = TRUE,
row.names = 1,
check.names = FALSE)
)Convert the matrix into a network object with the as.network() function, specifying that the network is bipartite and directed through the appropriate arguments.
davis1_net <- as.network(davis_mat,
# Should the network be interpreted as bipartite?
bipartite = TRUE,
# Should the edges be interpreted as directed?
directed = FALSE,
# Ignore edge values?
ignore.eval = FALSE,
# Optional edgeset constructor argument:
matrix.type = "incidence")
davis1_net Network attributes:
vertices = 32
directed = FALSE
hyper = FALSE
loops = FALSE
multiple = FALSE
bipartite = 18
total edges= 89
missing edges= 0
non-missing edges= 89
Vertex attribute names:
vertex.names
Edge attribute names:
NULL Note that you can use the is.bipartite() function to make sure the object is indeed a bipartite (two-mode) network.
is.bipartite(davis1_net)[1] TRUE4.4.1.2 Option 2: Importing Two-Mode Network Data as an Edge List
Let’s begin by importing an edge list and then check the first few rows with the head() command.
davis_el <- read.csv("data/davisedge.csv",
header = TRUE)
head(davis_el) Source Target Weight
1 EVELYN E1 1
2 EVELYN E2 1
3 EVELYN E3 1
4 EVELYN E4 1
5 EVELYN E5 1
6 EVELYN E6 1As you can see, the first column is the women and the second is the events they attended. This is how a two-mode edge list should be organized: the first mode will be whatever is represented in the first column and the second mode represented in the second.
To read a bipartite edge list to statnet use the as.network() function like before. Specify the matrix.type as edgelist, directed = TRUE, and bipartite = TRUE.
davis2_net <- as.network(davis_el,
matrix.type = "edgelist",
directed = FALSE,
bipartite = TRUE)
davis2_net Network attributes:
vertices = 32
directed = FALSE
hyper = FALSE
loops = FALSE
multiple = FALSE
bipartite = 18
total edges= 89
missing edges= 0
non-missing edges= 89
Vertex attribute names:
vertex.names
Edge attribute names:
Weight Let’s check the graph object by plotting it. Once again, you will have to specify the type of graph being evaluated by gplot(); to do so, set the gmode argument to twomode. Note that the women are colored red and events are colored red.
gplot(davis2_net,
gmode = "twomode",
usearrows = FALSE,
displaylabels = TRUE,
label.pos = 5,
label.cex = .6)
4.4.2 Plotting Two-Mode Networks
At this point, we have already plotted one bipartite network. Here we will compare a few plots using only the davis1_net network. Note that we need to tell gplot() that it is a two-mode network (with the argument gmode = "twomode"). Like before, we will compare layout algorithms side-by-side on the same row.
# Set graph parameters to 1 row and 3 columns
par(mfrow = c(1, 3))
# Store coordinates
coords_fr <- gplot.layout.fruchtermanreingold(davis1_net,
layout.par = NULL)
coords_kk <- gplot.layout.kamadakawai(davis1_net,
layout.par = NULL)
coords_cr <- gplot.layout.circle(davis1_net,
layout.par = NULL)
# Plot graphs
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_fr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
usearrows = FALSE)
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_kk,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
usearrows = FALSE)
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_cr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
usearrows = FALSE)
The default colors for statnet are blue and red, so if we want to assign different colors we can do so by creating a separate color vector.
# First, create a vector of length 18 with the value "light blue"
women <- rep("light blue", times = 18)
# Next, create a vector of length 14 with the value "yellow"
events <- rep("yellow", times = 14)
# Now, combine both into a single vector
color <- c(women, events)Now, replot the same networks as above. The only difference in the following commands from those above is that they use the stored coordinates and the color vector we just created.
par(mfrow = c(1, 3))
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_fr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = color,
usearrows = FALSE)
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_kk,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = color,
usearrows = FALSE)
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_cr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = color,
usearrows = FALSE)
Let’s calculate two-mode degree centrality and then assign the scores as actor attributes. First, let’s take a look at how to calculate node degree.
degree(davis1_net) [1] 16 14 16 14 8 8 8 6 8 8 8 12 14 16 10 4 4 4 6 6 12 8 16 16 20
[26] 28 24 10 8 12 6 6Note that you can assign that vector of scores as a vertex attributes.
davis1_net <- set.vertex.attribute(davis1_net,
attrname = "degree",
value = degree(davis1_net))
davis1_net Network attributes:
vertices = 32
directed = FALSE
hyper = FALSE
loops = FALSE
multiple = FALSE
bipartite = 18
total edges= 89
missing edges= 0
non-missing edges= 89
Vertex attribute names:
degree vertex.names
Edge attribute names:
NULL You can always call this attribute back.
get.vertex.attribute(davis1_net, "degree") [1] 16 14 16 14 8 8 8 6 8 8 8 12 14 16 10 4 4 4 6 6 12 8 16 16 20
[26] 28 24 10 8 12 6 6Plot graph with node size reflecting two-mode degree centrality. The degree scores are rescaled so that the vertices don’t overwhelm the graph.
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_fr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = color,
vertex.cex = get.vertex.attribute(davis1_net,
attrname = "degree")/10,
usearrows = FALSE)
4.4.3 Projecting (Folding) Two-Mode Networks into One-Mode Networks in statnet.
For this section, we will just use the davis1_net network ojbect.
4.4.3.1 Multiplying Matrices
We can transform the network into a one-mode network of the women by multiplying the matrix (not the graph) by its transpose in order to get a one-mode of the women-to-women.
First, let’s take a look at how to extract an adjacency matrix from the graph.
as.matrix.network.adjacency(davis1_net) E1 E2 E3 E4 E5 E6 E7 E8 E9 E10 E11 E12 E13 E14
EVELYN 1 1 1 1 1 1 0 1 1 0 0 0 0 0
LAURA 1 1 1 0 1 1 1 1 0 0 0 0 0 0
THERESA 0 1 1 1 1 1 1 1 1 0 0 0 0 0
BRENDA 1 0 1 1 1 1 1 1 0 0 0 0 0 0
CHARLOTTE 0 0 1 1 1 0 1 0 0 0 0 0 0 0
FRANCES 0 0 1 0 1 1 0 1 0 0 0 0 0 0
ELEANOR 0 0 0 0 1 1 1 1 0 0 0 0 0 0
PEARL 0 0 0 0 0 1 0 1 1 0 0 0 0 0
RUTH 0 0 0 0 1 0 1 1 1 0 0 0 0 0
VERNE 0 0 0 0 0 0 1 1 1 0 0 1 0 0
MYRNA 0 0 0 0 0 0 0 1 1 1 0 1 0 0
KATHERINE 0 0 0 0 0 0 0 1 1 1 0 1 1 1
SYLVIA 0 0 0 0 0 0 1 1 1 1 0 1 1 1
NORA 0 0 0 0 0 1 1 0 1 1 1 1 1 1
HELEN 0 0 0 0 0 0 1 1 0 1 1 1 0 0
DOROTHY 0 0 0 0 0 0 0 1 1 0 0 0 0 0
OLIVIA 0 0 0 0 0 0 0 0 1 0 1 0 0 0
FLORA 0 0 0 0 0 0 0 0 1 0 1 0 0 0Now, let’s generate a one-mode matrix of women-to-women relations based on shared participation in an event. To do so, we will multiply the adjacency matrix using the %*% operator times its transpose (t()).
davis_women_mat <- as.matrix.network.adjacency(davis1_net) %*%
t(as.matrix.network.adjacency(davis1_net))View the matrix:
davis_women_matRepeat the process, this time switch the order of the transposed matrix to generate an events-to-events matrix.
davis_events_mat <- t(as.matrix.network.adjacency(davis1_net)) %*%
as.matrix.network.adjacency(davis1_net)Finally, convert the new matrices into network objects.
davis_women_net <- as.network(davis_women_mat)
davis_events_net <- as.network(davis_events_mat)4.4.4 Plotting Projected One-Mode Networks
Now that we have extracted the one-mode networks, plot the two new graphs using gplot() and the additional arguments used previously.
par(mfrow = c(1, 2))
# Save coordinates
coords_women_kk <- gplot.layout.kamadakawai(davis_women_net,
layout.par = NULL)
coords_events_kk <- gplot.layout.kamadakawai(davis_events_net,
layout.par = NULL)
# Plot graphs
gplot(dat = davis_women_net,
gmode = "onemode",
coord = coords_women_kk,
label = network.vertex.names(davis_women_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "light blue",
usearrows = FALSE)
gplot(dat = davis_events_net,
gmode = "onemode",
coord = coords_events_kk,
label = network.vertex.names(davis_events_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "yellow",
usearrows = FALSE)
Resize the nodes by degree centrality. This time, we will not store the value as a vertex attribute.
par(mfrow = c(1, 2))
# Plot graphs
gplot(dat = davis_women_net,
gmode = "onemode",
coord = coords_women_kk,
label = network.vertex.names(davis_women_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "light blue",
vertex.cex = degree(davis_women_net, gmode = "graph")/10,
usearrows = FALSE)
gplot(dat = davis_events_net,
gmode = "onemode",
coord = coords_events_kk,
label = network.vertex.names(davis_events_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "yellow",
vertex.cex = degree(davis_events_net, gmode = "graph")/10,
usearrows = FALSE)
4.4.5 Saving Network Plots
Now, save plots of the two-mode network and the two one-mode networks.
png(file = "davis1.png", width = 4, height = 4, units = 'in', res = 300,
bg = "transparent")
gplot(dat = davis1_net,
gmode = "twomode",
coord = coords_fr,
label = network.vertex.names(davis1_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = color,
vertex.cex = get.vertex.attribute(davis1_net, attrname = "degree")/10,
usearrows = FALSE)
dev.off()
png(file = "daviswomen.png", width = 4, height = 4, units = 'in', res = 300,
bg = "transparent")
gplot(dat = davis_women_net,
gmode = "onemode",
coord = coords_women_kk,
label = network.vertex.names(davis_women_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "light blue",
vertex.cex = degree(davis_women_net, gmode = "graph")/10,
usearrows = FALSE)
dev.off()
png(file = "davisevents.png", width = 4, height = 4, units = 'in', res = 300,
bg = "transparent")
gplot(dat = davis_events_net,
gmode = "onemode",
coord = coords_events_kk,
label = network.vertex.names(davis_events_net),
label.col = "black",
label.cex = 0.6,
label.pos = 5,
vertex.col = "yellow",
vertex.cex = degree(davis_events_net, gmode = "graph")/10,
usearrows = FALSE)
dev.off()4.4.6 Saving Network Data
Once again, it doesn’t hurt to save the data that you’ve imported and created.
save(davis_el,
davis1_net,
davis2_net,
davis3_net,
davis_events_net,
davis_women_net,
file = "data/davis_statnet.RData")That’s all for statnet for now.