Bayesian Change Point Detection In Segmented Multi-Group Autoregressive Moving-Average Data For The Study Of COVID-19 In Wisconsin

Mr. Russell Latterman
University of Wisconsin-Milwaukee

Changepoint detection involves the discovery of abrupt fluctuations in population dynamics over time. We take a Bayesian approach to estimating points in time at which the parameters of an autoregressive moving average (ARMA) change, applying a Markov Chain Monte Carlo method. We specifically assume that data may originate from one of two groups. We provide estimates of all multi-group parameters of a model of this form for both simulated and real-world data sets. We include a provision to resolve the problem of confounding ARMA parameter estimates and variance of segment data. We apply our model to identify events that may have contributed to the 2020 and 2021 outbreaks of COVID-19 in Waukesha County, Wisconsin.

Advisor: Prof. David Spade

Committee Members:
Profs. Richard Stockbridge, Istvan Lauko, Chao Zhu, and Vytaras Brazauskas