Change point modeling for spatio-temporal ordinal data

We define a Bayesian model to:

1. identify changes in relationships with spatio-temporal ordinal data,
2. quantify the changes in time, and
3. account for spatio-temporal dependence.

Several challenges make such a model difficult, including the nature of ordinal data and estimating the many relationships of such data. We apply our model to COVID-19 transmission levels (as defined by the CDC) in New York State from January 20, 2020 to May 16, 2022.

Illustration of the estimated changing spatio-temporal dependence for various weeks during the pandemic.