Daily number of hospitalizations and deaths are key outcomes in quantifying the outbreak of infectious diseases. For the purposes of understanding the trend of the processes and the effect of observations from previous days, it may be useful to consider time series approaches for modeling the outcomes. Using such an approach, cointegration analysis may be employed to identify the long-run relationship between those multiple processes that are key to understanding trends in infectious disease such as hospitalization and death. As an alternative perspective, relationships between outcomes can be modeled through a shared latent stochastic error term; here, we propose a novel framework to study the underlying correlation between two time series processes through this method called joint modeling. In our Ontario Covid-19 study, a cointegration analysis utilizes statistical tests to identify the long-run relationship between the daily number of new hospitalizations six days prior and the daily number of new deaths in Ontario. Additionally, a joint autoregressive model provides a framework to model the underlying correlation between the processes.

Dexen Xi is a PhD candidate in the Statistical and Actuarial Sciences at Western University. His thesis is on statistical methods with a focus on joint outcome modeling and on methods for fire science.

AUTHORS: corresponding author: Dexen DZ. Xi, Department of Statistical and Actuarial Sciences, University of Western Ontario, London, Ontario, N6A 5B7, Canada; email: dxi@uwo.ca and C.B. Dean, Department of Statistical and Actuarial Science, University of Waterloo, Waterloo, Ontario, N2L 3G1, Canada; email:cdean@uwaterloo.ca.

Related publication: Xi D. D. Z., Dean, C.B., & Taylor, S.W. (2020). Modeling the duration and size of extended attack wildfires as dependent outcomes. Environmetrics. 31(e2619). https://doi.org/10.1002/env.2619