Presentation
Parameter Estimation and Spreading Model Selection in a Spatiotemporal Network Framework Using Approximate Bayesian Computation
Presenter
DescriptionABC-SMC is a computational method in Bayesian statistics that combines Approximate Bayesian Computation with Sequential Monte Carlo sampling. This method is ideal for stochastic, complex models where the likelihood function is intractable or computationally expensive to evaluate. We adapt Approximate Bayesian Computation based on the Sequential Monte Carlo sampling (ABC-SMC) method in the context of an individual-based network model for parameter estimation and network model selection. This work provides a general, flexible, and comprehensive framework to study epidemic processes at the individual level. We have used ABC-SMC in an individual-based heterogeneous network framework to understand the spatial distribution of the West Nile virus (WNV) in the United States. We propose distance dispersal kernels, which describe the probability of dispersal with respect to distances. We propose three types of distance dispersal kernels: 1) exponential, 2) power-law, and 3) power-law biased by flyway. We then estimate the disease parameters and compare these three distance kernels using incidence data with the ABC-SMC method. After conducting extensive simulations for the years 2014–2016, we observed that an adapted fat-tailed or power-law kernel, which includes long-distance links in specific directions, best describes the WNV human case data.
TimeTuesday, June 1716:00 - 16:30 CEST
LocationRoom 6.0D13
Event Type
Minisymposium
Applied Social Sciences and Humanities
Life Sciences
Computational Methods and Applied Mathematics