Frontier Problems in Modeling Dynamic Social Systems
Social dynamics – changes in the structure, existence or occurrence of social connections (e.g. sexual, friendship) or events (e.g. conversation, armed conflict) – are a core interest of social sciences (Comte, 1855; Weber, 1904; Marx and Engels, 1972). One key to understanding dynamic social systems is to understand the social networks that comprise these systems (Mayhew, 1984; Moreno, 1934). Modeling networks over time is of essential interest to understanding social systems and structure, from intimate sexual partnership ties (Lévi-Strauss, 1969; Morris and Kretzschmar, 1995; Hudson, 1993; Moody, 2002) to the interactions of large-scale organizations (Blau, 1970; Childers et al., 1971; Mayhew et al., 1972). Recently, advances in computing and statistical methodology have improved our ability to model relational dynamics. One such methodology is the exponential family models for modeling social networks (i.e. temporal exponential random graph models (TERGMs), see (Krivitsky and Handcock, 2014)). This thesis aims to improve our ability to model relational dynamics through a substantive exploration of a dynamic social system, a purely predictive imputation of edge dynamics among social media accounts, and an exploration of how a dynamic model handles temporal adjustments. I thus aim to address frontier problems in relational dynamics, to improve and extend our current models by both substantive application and methodology assessment.