Title: Spatial kriging for replicated anisotropic point processes
Authors: Daniel Gervini - University of Wisconsin-Milwaukee (United States) [presenting]
Abstract: In many applications, a temporal point process is observed at different spatial locations, and it is of interest to predict the temporal process at a new location. For example, in the Divvy bicycle-sharing system of the city of Chicago, bike checkout times can be seen as a temporal point process observed every day at different spatial locations, the bike stations, and it is of interest to predict bike demand at potential locations outside the grid, in order to determine where to best place new stations. A spatial kriging approach to this problem will be presented. Unlike common kriging methods that assume isotropy of the spatial field and would not be applicable to the Divvy data, which is clearly anisotropic, the proposed method uses the daily replications of the process at each site to nonparametrically estimate the spatial means and covariances and does not need the isotropy assumption.