This script, presented at the April 2015 Meetup, demonstrates how to carry out basic social network analysis (SNA) using the network and sna packages. Topics covered include preparing data for analysis, interpreting output, and generating social network graphs.
The script assumes no prior knowledge of SNA but does assume a basic working knowledge of R. Most code is commented with an explanation of its purpose.
After this demonstration, hopefully Meetup participants will have an understanding of how to begin and carry out a basic social network analysis in R.
# ---------------------------------
# work with network/relational data
# ---------------------------------
# Read in your data ------------------------------------------------
setwd('C:/Users/yt3f/ytai/Rusersgroup')
# We will use "gotRel.csv" here. Let's first take a look of the original file.
# Import an adjacency matrix using read.csv() function.
gotdf <- read.csv(file='gotRel.csv',header=TRUE,row.names=1,stringsAsFactors=FALSE)
gotdf # header=T and row.names=1 help to make it as a 10 by 10 matrix
gotmat <- as.matrix(gotdf) # convert it to a matrix form
# The "Network" package: "network" objects -------------------------------------------
# A network object is a R object that can store both relational data and its metadata.
# Here we use the "network" package to construct, describe, manipulate, and plot network objects.
#install.packages('network')
library(network)
# Using as.network.matrix() to convert the matrix (gotmat) into a network object,
# which is undirected and carries edge values named "weight"
gotmat
gotnet <- as.network.matrix(gotmat,directed=FALSE,ignore.eval=FALSE,names.eval="weight")
gotnet # quick description of the network
summary(gotnet) # more info of the network
network.edgecount(gotnet) # how many edges in the network
network.size(gotnet) # how many nodes in the network
plot(gotnet) # a quick plot of the network
plot(gotnet,displaylabels=TRUE) # put on node names
# In case you like to add or delete edges:
network.vertex.names(gotnet) # list the nodes
gotnet[2,3] # are Cersei and Ned connected?
gotnet[2:3,4:7] # subset of row 2 to 3. column 4 to 7
gotnet[-10,-10] # if you want to kill Sansa (from the graph, of course...)
gotnet[2,] # all of Cersei's ties, i.e., extract the Cersei row
# note that this doesn't display edge values
gotmat[2,]
as.sociomatrix(gotnet,attrname="weight") # retrieving edge values in a matrix form
as.sociomatrix(gotnet) # the default gives you a dichotomous one
gotnet[,] # the other way to do it
# plot the network with presenting edge values by line width
plot(gotnet,displaylabels=TRUE,edge.lwd="weight")
# Network and node attributes --------------------------------------------
# Let's take a look of the node attributes file
# import nodes attributes data
gotatt <- read.csv(file='gotAttri.csv',header=TRUE,stringsAsFactors=FALSE)
gotatt
# load the node atrributes into the network object, i.e., gotnet
gotnet %v% "id" <- gotatt$ID
gotnet %v% "name" <- gotatt$NAME
gotnet %v% "sex" <- gotatt$SEX
gotnet %v% "family" <- gotatt$FAMILY
gotnet %v% "status" <- gotatt$STATUS
gotnet %v% "appear" <- gotatt$APPEAR
gotnet
list.vertex.attributes(gotnet) # list node atrribute names
gotnet %v% "name" # retrieve the "name" attribute
# ---------------------------
# Some Basic Network Analysis
# ---------------------------
# use a build-in dataset
data(package="network") # datasets come with the network package
?flo
data(flo) # load the "flo" dataset
flo # take a look of it
is.matrix(flo)
#install.packages('sna')
library(sna)
# The "sna" package can handle many forms of network data, e.g., network object, adjacency matix,
# or a list of adjacency matrices. In other words, functions of this package can be used on
# various forms of data. We'll work with edgelist in the following example.
floel <- as.edgelist.sna(flo) # convert the flo matrix into a sna edgelist
head(floel)
attr(floel,"n") # how many nodes?
attr(floel,"vnames") # node names
as.sociomatrix.sna(floel) # convert it to a sna matrix
# Measure centrality ---------------------------------------
data(emon) # load the "emon" dataset
?emon # it contains 7 network objects
emon.tx <- emon$Texas # we'll only use the Texas one here
emon.tx # a directed network
# degree cnetrality
degree(emon.tx) # total degree of each node
degree(flo) # works for a matrix object as well
indg <- degree(emon.tx, cmode="indegree") # in-degree
indg
oudg <- degree(emon.tx, cmode="outdegree") # out-degree
oudg
# plot in-degree by out-degree
plot(x=indg,y=oudg,xlab="Indegree",ylab="Outdegree")
# Which GOT character is more central than others, in terms of degree centrality?
gotdg <- degree(gotnet)
max(gotdg) # the max value in gotdg
which(gotdg==12) # returns the position of the value=12
network.vertex.names(gotnet)
# betweeness and closeness centrality
bet <- betweenness(emon.tx)
bet # betweenness centrlity of each node
clo <- closeness(emon.tx)
clo # closeness centrlity of each node
# plot the relationship of closeness and betweenness
plot(x=clo, y=bet, xlab="Closeness", ylab="Betweenness")
# how about betweenness and closeness in our GOT network
gotbet <- betweenness(gotnet,gmode="graph") # "digraph" by default
gotclo <- closeness(gotnet,gmode="graph")
max(gotbet)
which(gotbet==18)
max(gotclo)
which(gotclo==0.75)
# --------------------
# Network Visulization
# --------------------
plot(gotnet) # default plot
plot(gotnet, displaylabels=TRUE) # add labels
plot(gotnet, displaylabels=TRUE,
vertex.cex=4, # expansion factor=>node size
vertex.border="red", # node border color
label.cex=0.8, # label size
label.col="white", # label color
label.pos=5) # label position, here we place labels at the vertex positions
# Of course we want to color the nodes by gender stereotype...
ncolor <- ifelse(gotatt$SEX=="F",yes="lightpink",no="lightblue")
ncolor # a list of color names assigned to each nodes
plot(gotnet,displaylabels=T,vertex.cex=3,vertex.col=ncolor,label.pos=5)
# or the "status" of them
ncolor2 <- ifelse(gotatt$STATUS=="alive",yes="green",no="antiquewhite3")
plot(gotnet,displaylabels=T,vertex.cex=3,vertex.col=ncolor2,label.pos=5)
# a directed network
plot(emon.tx)
par(mar=c(0,0,0,0)) # get rid of the large margin; the numerical vector indicating
# margin size c(bottom, left, top, right) in lines.
plot(emon.tx)
plot(emon.tx, displaylabels=TRUE)
# try a different layout
plot(emon.tx, displaylabels=TRUE,
vertex.cex=2,
label.cex=0.7,
label.col="midnightblue",
label.pos=5,
mode="circle") # vertex placement algorithm
?gplot.layout # more info of layouts
# interactive mode for tweaking
plot(emon.tx, displaylabels=TRUE,
vertex.cex=2,
label.cex=0.7,
label.col="midnightblue",
label.pos=5,
mode="circle",
interactive=TRUE) # interactive mode
# save a layout for reuse
Coords1 <- plot(emon.tx, displaylabels=TRUE,
vertex.cex=2,
label.cex=0.7,
label.col="midnightblue",
label.pos=5,
mode="circle",
interactive=TRUE)
head(Coords1) # the nodes coordinates
plot(emon.tx)
plot(emon.tx, coord=Coords1) # apply a saved layout
# more info of graph plotting
?gplot