Spatial Analysis of Nigeria’s Presidential Election Based on Geographically Weighted Regression

Abstract:

Geographically weighted regression (GWR) is a linear regression technique used to fit a regression equation to every observation in a dataset. In this study, both the global regression (multiple linear regression) and the GWR were calibrated for the 2019 Nigeria presidential election dataset, and diagnostics of each model were computed and compared. Experiments and analyses in the study were implemented in the R-environment, R-4.1.2. The GWR model outperforms the global regression model with an R^2 value of 0.776 exceeding that of the global regression of 0.513. The superiority of the GWR model is also confirmed by its much smaller RSS and AICc values (173.362 and 1372.8555 respectively), compared to those of the global regression (377.103 and 1662.316 respectively). The GWR model better fits the election dataset; it explains spatial variations in the dependent variable better than the global regression model does.