Multiscale Geographically Weighted Regression (MGWR)
This module provides functionality to calibrate multiscale (M)GWR as well as traditional GWR. It is built upon the sparse generalized linear modeling (spglm) module.
- GWR model calibration via iteratively weighted least squares for Gaussian, Poisson, and binomial probability models.
- GWR bandwidth selection via golden section search or equal interval search
- GWR-specific model diagnostics, including a multiple hypothesis test correction and local collinearity
- Monte Carlo test for spatial variability of parameter estimate surfaces
- GWR-based spatial prediction
- MGWR model calibration via GAM iterative backfitting for Gaussian model
- MGWR covariate-specific inference, including a multiple hypothesis test correction and local collinearity