sna
is a package containing a range of tools for social network analysis. Supported functionality includes node and graph-level indices, structural distance and covariance methods, structural equivalence detection, p* modeling, random graph generation, and 2D/3D network visualization (among other things).
Network data for sna
routines can (except as noted otherwise) appear in any of the following forms:
adjacency matrices (dimension N x N);
arrays of adjacency matrices, aka “graph stacks” (dimension m x N x N);
sna edge lists (see below);
sparse matrix objects (from the SparseM package);
network
objects (from the network package); or
lists of adjacency matrices/arrays, sparse matrices, and/or network
objects.
Within the package documentation, the term “graph” is used generically to refer to any or all of the above (with multiple graphs being referred to as a “graph stack”). Note that usage of sparse matrix objects requires that the SparseM package be installed. (No additional packages are required for use of adjacency matrices/arrays or lists thereof, though the network package, on which sna depends as of 2.4, is used for network objects.) In general, sna
routines attempt to make intelligent decisions regarding the processing of multiple graphs, but common sense is always advised; certain functions, in particular, have more specific data requirements. Calling sna
functions with inappropriate input data can produce “interesting” results.
One special data type supported by the sna package (as of version 2.0) is the sna edgelist. This is a simple data format that is well-suited to representing large, sparse graphs. (As of version 2.0, many - now most - package routines also process data in this form natively, so using it can produce significant savings of time and/or memory. Prior to 2.0, all package functions coerced input data to adjacency matrix form.) An sna edgelist is a three-column matrix, containing (respectively) senders, receivers, and values for each edge in the graph. (Unvalued edges should have a value of 1.) Note that this form is invariant to the number of edges in the graph: if there are no edges, then the edgelist is a degenerate matrix of dimension 0 by 3. Edgelists for undirected graphs should be coded as fully mutual digraphs (as would be the case with an adjacency matrix), with two edges per dyad (one (i,j) edge, and one (j,i) edge). Graph size for an sna edgelist matrix is indicated by a mandatory numeric attribute, named "n"
. Vertex names may be optionally specified by a vector-valued attribute named "vnames"
. In the case of two-mode data (i.e., data with an enforced bipartition), it is possible to indicate this status via the optional "bipartite"
attribute. Vertices in a two-mode edgelist should be grouped in mode order, with "n"
equal to the total number of vertices (across both modes) and "bipartite"
equal to the number of vertices in the first mode.
Direct creation of sna edgelists can be performed by creating a three-column matrix and using the attr
function to create the required "n"
attribute. Alternately, the function as.edgelist.sna
can be used to coerce data in any of the above forms to an sna edgelist. By turns, the function as.sociomatrix.sna
can be used to convert any of these data types to adjacency matrix form.
To get started with sna
, try obtaining viewing the list of available functions. This can be accomplished via the command library(help=sna)
.