Tweeting #RamNavami: A Comparison of Approaches to Analyzing Bipartite Networks

Bipartite networks, also known as two-mode networks or affiliation networks, are a class of networks in which actors or objects are partitioned into two sets, with interactions taking place across but not within sets. These networks are omnipresent in society, encompassing phenomena such as student-teacher interactions, coalition structures, and international treaty participation. With growing data availability and proliferation in statistical estimators and software, scholars have increasingly sought to understand the methods available to model the data generating processes in these networks. This article compares three methods for doing so: (1) Logit; (2) the bipartite Exponential Random Graph Model (ERGM); and (3) the Relational Event Model (REM). This comparison demonstrates the relevance of choices with respect to dependence structures, temporality, parameter specification, and data structure. Considering the example of Ram Navami, a Hindu festival celebrating the birth of Lord Ram, the ego network of tweets using #RamNavami on April 21, 2021 is examined. The results of the analysis illustrate that critical modeling choices make a difference in the estimated parameters and the conclusions to be drawn from them.