Geographically weighted regression (GWR) is a regional linear regression model which involves the consideration of spatial autocorrelation. It is popularly used in recent years for its better fitness than traditional OLS in explaining spatially varying relationship. The lab contains three parts. The first part is a general description of GWR, introducing its fundamental principles and advantages. The second part conducts a case study of GWR in Vancouver children’s social skill scores. The GWR result is analyzed and interpreted with the reference of a grouping analysis. OLS model is also built to compare the model fitness between the two regression models. The third part discusses how GWR can be applied in a broader environmental and socio-economic context.

In this lab, we explored the relationship between a child’s social skill ability and several variables related to the child and their neighborhood in the city of Vancouver. We found there were three most important parameters: gender, income, and language that would affect a child’ social skill ability. We modeled both global Ordinary least squares and local Geographically weighted regression by using the three parameters as explanatory variables to predict the social skill ability. And the result proofed that the better fitness of GWR than the OLS.

Figure 1. Grouping Analaysis Result