Social Network Analysis Using R

R and Its Community

R is one of the outstanding successes of the free software movement. The code is open for inspection, editing, and redistribution without restriction. A New York Times article published January 7, 2009, speculated that R was becoming a lingua franca of statistics and data science.

The R framework is modular; most of the work is done by functions contained in packages. It is easy for new R users to overlook the difference between the base of R (the software distributed by the R Core Team, which includes 30 packages) and packages provided by the R community, which now number more than 10,000 (taking together the Comprehensive R Archive Network as well as smaller repositories such as Bioconductor, R-Forge, and GitHub). The openness to addition of packages is a significant part of the explanation for R’s growth.

Another reason that R is becoming the lingua franca of statistics is that R can absorb functions written in fast, low-level programming languages such as C and Fortran. Experts may prefer to write in C++, for example. R packages that incorporate those functions often appear.

The base R distribution does not include tools for SNA. The general purpose social network [Page 1544]frameworks considered here are found in packages igraph, statnet, and graph. Depending on the researcher’s taste and needs, any or all of these may be useful. In combination with tools in R base, each one of these is able to handle the following:

• importation of data,
• description of networks (summarize connections among individuals),
• visualization (plotting and interaction with graphic displays), and
• Simulation of artificial networks.

Users can expect differing degrees of difficulty when using these packages. One package will offer nicer plots, but at the expense of more difficult data preparation, for example. Researchers will have to pick and choose among the functions offered by different packages. There are significant stylistic differences among packages and it is not always easy to navigate among them. The packagers are aware of these concerns and make frequent revisions. As a result, many blogs and tutorials about SNA are outdated. Books published as recently as 2014 describe functions in packages such as igraph, which no longer exist or are being phased out.

There is no single R package that can handle all of the more advanced needs of social network researchers. The statnet suite is the closest to that objective. It links together 14 separate R packages; especially noteworthy are the data importers in network, network connectivity analysis in sna, and exponential random graph models (ERGM) in ergm.

...