/gwr

This is a repository for the geographically-weighted regression submodule of the Python Spatial Analysis Library

Primary LanguageJupyter NotebookBSD 2-Clause "Simplified" LicenseBSD-2-Clause

Geographically Weighted Regression

This module provides geographically weighted regression functionality. It is built upon the sparse generalized linear modeling (spglm) module.

Features

The gwr module currently features

  • gwr model estimation via iteratively weighted least squares for Gaussian, Poisson, and binomial probability models.
  • gwr bandwidth selection via golden section search
  • gwr-specific model diagnostics, including a multiple hypothesis testing correction
  • gwr-based spatial prediction

Future Work

  • Additional probability models (gamma, negative binomial)
  • Tests for spatial variability
  • Multi-scale gwr