Geographically Weighted Regression (GWR) is a method of spatial analysis that examines how the relationships between variables can vary geographically. The method involves fitting a series of distance weighted regression models to spatial subsets of a source data set and then pooling the results - a process which can take a long time to run when the size of the dataset becomes large.
Fortunately, the development of the UK’s National Grid Service and e-science programmes offers an environment to run a parallel implementation of GWR that greatly decreases the time required to obtain the results. Here we report on an ESRC NCeSS funded project that permits us to run GWR models using desktop R (an open source environment for computing and statistics).
|
