This line of inquiry investigates how scale can be better captured in multivariate statistical models, such as multiscale geographically weighted regression (MGWR) and generalized additive spatial smoothing (GASS). This work entails developing algorithms for both the inference of spatial relationships and the prediction of spatial observations. It also seeks to scale these methods, which are frequently computationally cumbersome, so that they can be used at higher resolutions (i.e., within cities) and for larger scopes (i.e., national and global). Ongoing research includes several applications and developing an open source Python implementation of MGWR.

This work is supported by the National Science Foundation (#2117455 and #1758786)