Smooth marginalized particle filters for dynamic network effect models

We propose the dynamic network effect (DNE) model for the study of high-dimensional multivariate time series data. Cross-sectional dependencies between units are captured via one or multiple observed networks and a low-dimensional vector of latent stochastic network effects. The parameter-driven, nonlinear state-space model requires simulation-based filtering and estimation, for which we suggest to use the smooth marginalized particle filter (SMPF). In a Monte Carlo simulation study, we demonstrate the SMPF's good performance relative to benchmarks, particularly when the cross-section dimension is large and the network is dense. An empirical application on the spread of the COVID-19 pandemic through international travel networks illustrates the usefulness of our method.