Search
mgwr

Multiscale Geographically Weighted Regression (MGWR)

Build Status Documentation Status PyPI version

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.

Features

  • 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