Title: Wild bootstrap for spatio-temporal data
Authors: Daisuke Kurisu - Tokyo Institute of Technology (Japan) [presenting]
Kengo Kato - Cornell University (United States)
Abstract: A novel bootstrap method is introduced for spatial and spatio-temporal data. For this, we derive Gaussian approximations and propose the spatial wild bootstrap (SWB) for sample means of a random field observed at a finite number of locations in a sampling region. In particular, we give Gaussian and bootstrap approximations for probabilities that the normalized sample means of discretely observed random field hit hyperrectangles. Additionally, we show that our results can be applied to a wide class of multivariate Levy driven moving average random fields and discuss multiple temporal change point tests for spatio-temporal data.