ergm (version 3.1.3)

plot.network.ergm: Two-Dimensional Visualization of Networks

Description

plot.network.ergm produces a simple two-dimensional plot of the network object x. A variety of options are available to control vertex placement, display details, color, etc. The function is based on the plotting capabilities of the network package with additional pre-processing of arguments. Some of the capabilites require the latentnet package. See plot.network in the network package for details.

Usage

## S3 method for class 'ergm':
plot.network(x,
    attrname=NULL,
    label=network.vertex.names(x),
    coord=NULL,
    jitter=TRUE,
    thresh=0,
    usearrows=TRUE,
    mode="fruchtermanreingold",
    displayisolates=TRUE,
    interactive=FALSE,
    xlab=NULL,                                      
    ylab=NULL,
    xlim=NULL,
    ylim=NULL,
    pad=0.2,
    label.pad=0.5,
    displaylabels=FALSE,
    boxed.labels=TRUE,
    label.pos=0,
    label.bg="white",
    vertex.sides=8,
    vertex.rot=0,
    arrowhead.cex=1,
    label.cex=1,
    loop.cex=1,
    vertex.cex=1,
    edge.col=1,
    label.col=1,
    vertex.col=2,
    label.border=1,
    vertex.border=1,
    edge.lty=1,
    label.lty=NULL,
    vertex.lty=1,
    edge.lwd=0,
    label.lwd=par("lwd"),
    edge.len=0.5,
    edge.curve=0.1,
    edge.steps=50,
    loop.steps=20,
    object.scale=0.01,
    uselen=FALSE,
    usecurve=FALSE,
    suppress.axes=TRUE,
    vertices.last=TRUE,
    new=TRUE,
    layout.par=NULL,
    cex.main=par("cex.main"), 
    cex.sub=par("cex.sub"),
    seed=NULL,
    latent.control=list(maxit=500,
                        trace=0,
                        dyadsample=10000,
                        penalty.sigma=c(5,0.5), 
                        nsubsample=200),
    colornames="rainbow",
    verbose=FALSE, 
    latent=FALSE, 
    ...)

Arguments

x

an object of class network.

attrname

an optional edge attribute, to be used to set edge values.

label

a vector of vertex labels, if desired; defaults to the vertex labels returned by network.vertex.names.

coord

user-specified vertex coordinates, in an NCOL(dat)x2 matrix. Where this is specified, it will override the mode setting.

jitter

boolean; should the output be jittered?

thresh

real number indicating the lower threshold for tie values. Only ties of value >thresh are displayed. By default, thresh=0.

usearrows

boolean; should arrows (rather than line segments) be used to indicate edges?

mode

the vertex placement algorithm; this must correspond to a network.layout function. These include "latent", "latentPrior", and "fruchtermanreingold".

displayisolates

boolean; should isolates be displayed?

interactive

boolean; should interactive adjustment of vertex placement be attempted?

xlab

x axis label.

ylab

y axis label.

xlim

the x limits (min, max) of the plot.

ylim

the y limits of the plot.

pad

amount to pad the plotting range; useful if labels are being clipped.

label.pad

amount to pad label boxes (if boxed.labels==TRUE), in character size units.

displaylabels

boolean; should vertex labels be displayed?

boxed.labels

boolean; place vertex labels within boxes?

label.pos

position at which labels should be placed, relative to vertices. 0 results in labels which are placed away from the center of the plotting region; 1, 2, 3, and 4 result in labels being plac

label.bg

background color for label boxes (if boxed.labels==TRUE); may be a vector, if boxes are to be of different colors.

vertex.sides

number of polygon sides for vertices; may be given as a vector or a vertex attribute name, if vertices are to be of different types.

vertex.rot

angle of rotation for vertices (in degrees); may be given as a vector or a vertex attribute name, if vertices are to be rotated differently.

arrowhead.cex

expansion factor for edge arrowheads.

label.cex

character expansion factor for label text.

loop.cex

expansion factor for loops; may be given as a vector or a vertex attribute name, if loops are to be of different sizes.

vertex.cex

expansion factor for vertices; may be given as a vector or a vertex attribute name, if vertices are to be of different sizes.

edge.col

color for edges; may be given as a vector, adjacency matrix, or edge attribute name, if edges are to be of different colors.

label.col

color for vertex labels; may be given as a vector or a vertex attribute name, if labels are to be of different colors.

vertex.col

color for vertices; may be given as a vector or a vertex attribute name, if vertices are to be of different colors.

label.border

label border colors (if boxed.labels==TRUE); may be given as a vector, if label boxes are to have different colors.

vertex.border

border color for vertices; may be given as a vector or a vertex attribute name, if vertex borders are to be of different colors.

edge.lty

line type for edge borders; may be given as a vector, adjacency matrix, or edge attribute name, if edge borders are to have different line types.

label.lty

line type for label boxes (if boxed.labels==TRUE); may be given as a vector, if label boxes are to have different line types.

vertex.lty

line type for vertex borders; may be given as a vector or a vertex attribute name, if vertex borders are to have different line types.

edge.lwd

line width scale for edges; if set greater than 0, edge widths are scaled by edge.lwd*dat. May be given as a vector, adjacency matrix, or edge attribute name, if edges are to have different line widths.

label.lwd

line width for label boxes (if boxed.labels==TRUE); may be given as a vector, if label boxes are to have different line widths.

edge.len

if uselen==TRUE, curved edge lengths are scaled by edge.len.

edge.curve

if usecurve==TRUE, the extent of edge curvature is controlled by edge.curv. May be given as a fixed value, vector, adjacency matrix, or edge attribute name, if edges are to have different levels of curvature.

edge.steps

for curved edges (excluding loops), the number of line segments to use for the curve approximation.

loop.steps

for loops, the number of line segments to use for the curve approximation.

object.scale

base length for plotting objects, as a fraction of the linear scale of the plotting region. Defaults to 0.01.

uselen

boolean; should we use edge.len to rescale edge lengths?

usecurve

boolean; should we use edge.curve?

suppress.axes

boolean; suppress plotting of axes?

vertices.last

boolean; plot vertices after plotting edges?

new

boolean; create a new plot? If new==FALSE, vertices and edges will be added to the existing plot.

layout.par

parameters to the network.layout function specified in mode.

cex.main

Character expansion for the plot title.

cex.sub

Character expansion for the plot sub-title.

seed

Integer for seeding random number generator. See set.seed.

latent.control

A list of parameters to control the latent and latentPrior models, dyadsample determines the size above which to sample the latent dyads; see ergm and

colornames

A vector of color names that can be selected by index for the plot. By default it is colors().

verbose

logical; if this is TRUE, we will print out more information as we run the function.

latent

logical; use a two-dimensional latent space model based on the MLE fit. See documentation for ergmm() in latentnet.

...

additional arguments to plot.

Value

• None.

Requires

mva

Details

plot.network is a version of the standard network visualization tool within the sna package. By means of clever selection of display parameters, a fair amount of display flexibility can be obtained. Network layout -- if not specified directly using coord -- is determined via one of the various available algorithms. These are (briefly) as follows:

1. latentPrior: Use a two-dimensional latent space model based on a Bayesian minimum Kullback-Leibler fit. See documentation forlatent()inergm.
2. random: Vertices are placed (uniformly) randomly within a square region about the origin.
3. circle: Vertices are placed evenly about the unit circle.
4. circrand: Vertices are placed in a ``Gaussian donut,'' with distance from the origin following a normal distribution and angle relative to the X axis chosen (uniformly) randomly.
5. eigen,princoord: Vertices are placed via (the real components of) the first two eigenvectors of:
   1. eigen: the matrix of correlations among (concatenated) rows/columns of the adjacency matrix
   2. princoord: the raw adjacency matrix.
6. mds,rmds,geodist,adj,seham: Vertices are placed by a metric MDS. The distance matrix used is given by:
   1. mds: absolute row/column differences within the adjacency matrix
   2. rmds: Euclidean distances between rows of the adjacency matrix
   3. geodist: geodesic distances between vertices within the network
   4. adj:$(\max A)-A$, where$A$is the raw adjacency matrix
   5. seham: structural (dis)equivalence distances (i.e., as persedistin the packagesna) based on the Hamming metric
7. spring,springrepulse: Vertices are placed using a simple spring embedder. Parameters for the embedding model are given byembedder.params, in the following order: vertex mass; equilibrium extension; spring coefficient; repulsion equilibrium distance; and base coefficient of friction. Initial vertex positions are in random order around a circle, and simulation proceeds -- increasing the coefficient of friction by the specified base value per unit time -- until ``motion'' within the system ceases. Ifspringrepulseis specified, then an inverse-cube repulsion force between vertices is also simulated; this force is calibrated so as to be exactly equal to the force of a unit spring extension at a distance specified by the repulsion equilibrium distance.

References

Wasserman, S., and Faust, K. (1994). ``Social Network Analysis: Methods and Applications.'' Cambridge: Cambridge University Press.

See Also

plot

Examples

data(florentine)
plot(flomarriage)  #Plot the Florentine Marriage data
plot(network(10))  #Plot a random network
plot(flomarriage,interactive="points")