Abstract by Zoe Gibbs
Using Bayesian Spatiotemporal Modeling to Understand Mortality Rates in the United States
Life in Seattle, Boston, and Dallas and the people who live there are different in many ways. Some of those are easily measurable (ethnicity, age, income, etc.) while others are not (stress levels, empathy, diet, exercise, etc.). Conversely, people in Dallas are relatively similar to those in Fort Worth (just don’t say that to anyone who lives there). In our project, we explore the spatial relationships between locations and build models to better predict both mortality rates and mortality improvement. We borrow strength from those locations which are close to each other but build flexible models which allow for the mortality rates to be very different between locations which are far apart. Specifically, we use Bayesian conditional autoregressive (CAR) models to understand the spatial and temporal variation of mortality by county in the United States.