Analyses of human responses to Winter storm Kai using the GWR model

This study examines differences of human responses to Winter storm Kai by employing the Geographically Weighted Regression (GWR) model. This study finds that Winter storm Kai exerts a different impact on human responses according to regions and periods. For instance, New York places first in the pre-winter storm week, whereas Washington takes first in the winter storm week. Second, the GWR model shows a lower estimator of prediction error and higher adjusted coefficient of determination than the Ordinary least squares (OLS), which means that the GWR model shows a better performance for the human responses to Winter storm Kai than OLS. Third, demographic variables play a different role in human responses. For example, while gender and race variables demonstrate the positive value for the number of tweets, age variables reveal the negative value for it. Fourth, the coefficient values of explanatory variables are differentiated by regions. For instance, race, education, and local R2 reveal the high value in Washington, whereas gender and age exhibit the high value in California.