Local Statistical Modeling

Catalog description

Understands the processes that generate the data we observe in the real world. If these processes vary over space, we term this spatial non-stationarity and traditional global models are no longer applicable. A set of local spatial models has been developed to examine spatial non-stationarity and one of the most widely used of these is Geographically Weighted Regression (GWR) and its variants. Investigates GWR and the wider context of spatial non-stationarity. Topics include: setting the scene; introduction to GWR; discussion of context; further issues in GWR; software for GWR; workshop on GWR 4; model selection in GWR; inference in GWR; semi-parametric GWR and multiscale GWR (MGWR); software for multiscale GWR; inference for MGWR; multicollinearity and GWR; some myths about GWR; big models; an example of the 2016 U.S. presidential election. Examination is by project (60%) and two presentations (40%).