A Novel Point Process Model for COVID-19: Multivariate Recursive Hawkes Process

Sat, 01 Jan 2022 00:00:00 +0000

The multivariate recursive Hawkes process extends self-exciting point-process models in two directions needed for epidemic data: transmission can interact across multiple populations, and the productivity of events can vary recursively rather than remaining fixed. This gives the model enough flexibility to represent heterogeneous and evolving COVID-19 transmission patterns.

The chapter establishes existence of the counting process and derives its first two moments. It also develops expectation-maximization procedures for parametric and semiparametric specifications, together with a reconstruction algorithm for observations whose event times are reported only coarsely.

Synthetic experiments validate the estimators, and applications to real COVID-19 data illustrate how the model recovers cross-population excitation and time-varying transmission effects from imperfect public-health records.

Hawkes Processes | Bohan Chen's Personal Webpage

A Novel Point Process Model for COVID-19: Multivariate Recursive Hawkes Process