 |
Center for Nonlinear Studies |
In an ever-more interconnected world, infectious diseases can travel around the world in days, posing a significant threat to public health and economic stability. In response to this trend, there is an increased demand for predictive models that can help public health officials assess the situation and develop timely mitigation strategies. Detailed, large-scale, stochastic, disease simulation capabilities have been developed to meet this challenge. However, due to imprecise knowledge about input parameters such as disease transmission rates and complex population dynamics, small perturbations on the system can have unpredictable and far-reaching effects. In this talk, I will present a method for characterizing the propagation of uncertainty through a model. Specifically, I will discuss a statistical emulation method that combines techniques from Gaussian process regression and polynomial chaos to account for multiple types of uncertainty. In addition, I will describe an ensemble Kalman filter and show how multiple implementations can lead to a wide variety of predicted results. Finally, I will discuss the relation between model emulation and data assimilation.
Host: Humberto C Godinez