Mathematical Models of Epidemics: Aggregating Stochastic Dynamics on Random Graphs

Meeting Details

For more information about this meeting, contact Donna Cepullio, Kristin Berrigan, Jessica Conway, Carina Curto, Andrew Belmonte, Timothy Reluga, Wenrui Hao, Leonid Berlyand, Vladimir Itskov.

Speaker: Grzegorz Rempala (Host: Wenrui Hao), Ohio State Univeresity

Abstract: Motivated by the classical Susceptible-Infected-Recovered (SIR) epidemic models and the data from current COVID-19 pandemic, we consider a class of stochastic dynamical systems (SDSs) evolving on random graphs. We show that the dynamics of such SDSs may be approximately described in terms of implicit survival functions and certain random measures. This survival interpretation allows us to employ tools from statistical theory of survival analysis to address various issues with data collection and statistical inference in classical SIR models. It also offers an alternative to more standard statistical methods based on the theory of hidden Markov processes. In particular, we propose and numerically validate a statistical inference procedure for SDS-likelihoods that is relying on observed marginal likelihoods generated by typically epidemic curves. Only a slightly more complected SDS model was successfully used by the state of Ohio to predict the amount of state COVID-19 burden in the early months of the 2020 pandemic. If time permits, I will also briefly outline the main ideas behind that specific model.

---

Room Reservation Information

Room Number: Zoom

Date: 03/30/2021

Time: 1:30pm - 2:30pm