Formation and Coevolution in Unsolicited Network Data

• Activity Number: 33
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Tyler H. McCormick*+ and Richard Li and Joshua Blumenstock
• Companies: University of Washington and University of Washington and University of Washington
• Keywords: self-exciting process ; social network

Abstract

Statistical methods for social network data generally model an observed graph representing the presence/absence of a relationship between all observed individuals (e.g. reporting an individual as a friend on a survey or claiming a "Friend" on Facebook). Emerging sources of unsolicited network data, such as telephone call logs or web histories, record the manifestations of social network structure but do not contain an explicitly defined graph. We develop a new statistical model designed to capture two processes. Using data with two types of electronic communications, we first propose a model that uses a mutually-exciting point process model to describe network formation. Next, we examine a new type of network process, known as coevolution, where two networks form simultaneously and the structure in one network impacts the other. Our data consist of cellphone communications and mobile money (a service for transfering funds via SMS text) in an east African nation. We observe the expansion of the cell phone network and the creation (and subsequent expansion) of mobile money transfers.