Multiscale Multivariate Spatio-temporal Bayesian Modeling of Covid-19 dynamics

Music City Center 

During the Covid-19 pandemic many attempts were made to model state or national level time series of case counts or mortality. To a much lesser extent, some effort was focused on spatial aspects of the pandemic and there are a few examples of spatio-temporal modeling at finer spatial scales. It is clear that spatial aspects are important in pandemic spread at a variety of scales (both spatial and temporal). Usually, the different scales are modelled separately. However there could easily be shared impact of linkages between levels. In this talk we outline the basic linkage models which can be set up across spatial scales and their potential linkages (mainly ) via random effects.
We give an example of modeling Covid19 mortality where weekly data is available now for 173 week periods at county level and state level in the US. A simple example of one state with associated counties is given. At the state level we have time series models and at the county level spatio-temporal modeling. Linkage is by shared effects. In principal, spatio-temporal models at state level can also be used. Comparison of models with and without linkage is made.
The approach can be extended to multivariate multiscale modeling for Covid19 data whereby case incidence, mortality and hopitalizations could be examined, or indeed the distribution of related infectious disease synchronous with Covid-19 such as RSV and influenza.

Bayesian

Multi-scale

Infectious disease

modeling

Space-time

Multivariate