Title: Quickest Change Detection for Monitoring Pandemics
Abstract: The problem of efficiently detecting changes in stochastic systems and time series, often referred to as the quickest change detection (QCD) problem, arises in various branches of science and engineering. It is assumed that the observations of the system undergo a change in distribution in response to a change or disruption in the environment. The observations are obtained sequentially, and if the state changes from the normal state, then it is of interest to detect this change as soon as possible, subject to false alarm constraints, and take any necessary action in response to the change. The goal in this talk is to explore how quickest change detection can be used in detecting the onset of a new wave of an existing pandemic using test positivity rate data.
Parametric and non-parametric approaches to constructing useful tests for pandemic monitoring will be discussed. In both approaches, the pre-change distribution is assumed to be stationary, while the post-change distribution is allowed to be non-stationary. In the parametric setting, the pre-change distribution is assumed to be known, while the post-change distribution is allowed to have parametric uncertainty. In the non-parametric approach, only knowledge of the pre-change mean and variance is assumed, and the post-change distributions are characterized simply as having mean greater than some pre-specified threshold, which is larger than the pre-change mean. New asymptotic theories are developed for both settings. Both approaches are shown to be effective in monitoring pandemics based on COVID-19 test positivity rate data.
(This is joint work with Yuchen Liang and Alexander Tartakovsky.)
