COLLOQUIUM Department of Computer Science and Engineering University of South Carolina Connecting the Dots: Applications of Exponential Random Graph Models to the Prediction of Tie/Arc/Edge Locations in Networks/Digraphs/Graphs John Skvoretz Department of Sociology University of South Carolina Date: October 25, 2002 (Friday) Time: 3:30-4:30PM Place: Swearingen 2A21 Abstract Recent advances in statistical models for networks, exponential random graph models, are outlined and applied to the question of predicting the presence of a tie between two nodes i and j in a network. These models allow the presence/absence of a tie to be affected by the local neighborhood of ties surrounding a particular couple, i and j, and by the attribute similarity or dissimilarity of i and j. Arenas for application include the Polish political opposition movement, the US Senate, and the Paris Metro. John Skvoretz is a Carolina Distinguished Professor of Sociology. He has undergraduate degrees in both Mathematics and Sociology and completed his PhD at the University of Pittsburgh in mathematical sociology. Research interests include the development and application of formal theoretical methods, such as simulation, stochastic processes, and statistical modeling and analysis, to a variety of sociological questions. Most recently, his research has focused on network analysis and modeling, including such issues as the development and testing of random and biased net models, diffusion of disease and information, the small world problem, population differentiation and network structure, and the application of exponential random graph models to the analysis of social networks.