Advanced Social Network Analysis and Visualization with R

Social network analysis benefits greatly from computational tools, and the R statistical language is a suitable environment to describe, visualize, and analyze network datasets of various sizes.  Social network analysis and network science is an area where intensive learning across disciplines is happening.  This course make new ideas from many disciplines – statistical physics, ecology, computer science – available to the social scientist, through libraries in R, and via the interactive possibilities of R notebooks available in R studio. We will use libraries such as ‘network’, ‘sna’, and ‘igraph’.  

The course is a good choice for those not taking the first week course (Social Networks: Theoretically Informed Analysis with UCINET), as we will re-visit the main concepts covered there briefly. In this case an awareness of the basic concepts of social network analysis is a plus.

This course is also a good choice to those who took the social network course in the first week using UCINET.  The course adds scalability of the R framework not present in UCINET (to analyze larger networks, many networks, or fork the analytic strategy to various alternative methods and indices).  The course also adds multiple approaches to hypothesis testing and comparisons to random baseline null hypotheses, sophisticated visualizations, and a broader range of algorithm choice.  The course also adds outlook to methods from network science, not present in UCINET (such as the replication of the Watts-Strogatz small world rewiring experiment) .

The first session will introduce R and programming basics, data import, and network descriptives.  We will discuss object types and programming  constructs relevant to network analysis. Various ways to import data to R will be explored. We will also learn about the generation of random graphs with diverse methods.  We will learn about various data structures – sociomatrices, edgelists, nodelists – and ways to transform data among these formats. We will import data from varous sources – Excel sheets, CSV and other delimited text formats, UCINET and other special data formats.

The second session is about what makes social networks distinctive: a strong tendency to closure and positive degree correlation (assortativity).  We will compare human social networks to animal social networks (and to decidedly not social network systems) to test the idea that animal social networks are indeed social.  We will discuss various centrality measures, and also centralization at the graph level. We will compare centrality measures, and discuss the theoretical imagery behind these measures, with detailed explanation of the formulas for the various indices.  

In the third session we will introduce measures for brokerage (structural autonomy), and also various brokerage roles at the boundaries of groups.  We will use rewirings to generate the random baseline expectation for the relative frequency of brokerage roles, and diagnose an organizational network of a thousand workers to detect bottlenecks in information flow in between divisions.

The fourth session looks at blocklmodeling and cohesive community detection, with an emphasis on goodness of fit measures, and comparisons with random baselines.  We will compare several methods of community detection, and discuss underlying assumptions about the nature of network data and processes of tie formation.

The last session is dedicated to advanced visualizations of network data. We discuss layout algorithms, and parameter choices for these – both for small and large graphs.  We also introduce  node density heat maps for extremely large graphs. We learn to convey information on graphs using node size and color, edge size and color. We discuss graph creation and export for various purposes: research article, PowerPoint, poster, interactive online presentation.

Basic knowledge of statistics and probability (probability density functions, tests of statistical significance, ordinary least squares estimator). Basic knowledge of data entry and manipulation (for example using Excel sheets, CSV, TXT data formats). 

A basic knowledge of R (objects, vectors, matrices) and awareness of the basic concepts in programming (for and while loops, if-else conditional statements) are a plus, but not required. These basics will be covered briefly, to the extent that enables independent learning.