Po-Ling Loh, University of Wisconsin

Two inference problems for network contagion
Dec 10, 2018, 4:30 pm5:30 pm
101 - Sherrerd Hall
Event Description

Abstract: We present two problems involving statistical inference for mathematical models of contagion spreading over a fixed network. The first problem concerns hypothesis testing for the underlying graph over which the disease is spreading, when we only observe the infection states of individual nodes after a single epidemic outbreak. We present a permutation test that is valid under appropriate conditions on the homogeneity of the spreading parameters and assumptions regarding the symmetry groups of the graphs involved in the null and alternative hypotheses. The second problem concerns parameter estimation for a similar type of contagion model, which incorporates covariate information on each of the edges of the graph. In this setting, we assume the structure of the graph is known, and we also know the order in which nodes contract the disease from their infected neighbors. We derive consistency and asymptotic normality of the maximum likelihood estimator, which may be obtained via convex optimization. This is joint work with Justin Khim (UPenn).

Bio: Po-Ling Loh is an assistant professor in the ECE department at the UW-Madison, with a secondary appointment in the statistics, computer science, and industrial and systems engineering departments. From 2014-2016, Po-Ling was an assistant professor in the statistics department at the Wharton School at the University of Pennsylvania. Po-Ling received an MS in computer science and a PhD in statistics from Berkeley in 2013 and 2014, and a BS in math with a minor in English from Caltech in 2009. She was the recipient of the 2014 Erich L. Lehmann Citation from the Berkeley statistics department for an outstanding PhD dissertation in theoretical statistics, and a best paper award at the NIPS conference in 2012. Po-Ling is a recipient of an NSF CAREER award in statistics.

Event Category
ORFE Department Colloquia