GitHub

SpatialEconometrics.jl

A comprehensive spatial econometrics package for Julia, providing maximum likelihood estimation of spatial regression models, panel data methods, diagnostics, and weight matrix construction.

Features

  • Cross-sectional models: SAR, SDM, SEM, SARAR via ML with exact or Chebyshev log-determinant
  • Panel models: Static and dynamic spatial panels with FE/RE, Lee-Yu bias correction, Direct QML
  • GS2SLS: Generalized spatial two-stage least squares with IV/endogenous regressors
  • System GMM: Arellano-Bond / Blundell-Bond for dynamic panels
  • GWR & MGWR: Geographically weighted regression with adaptive/fixed bandwidths and KDTree
  • Gravity models: PPML and Negative Binomial with origin/destination/bilateral fixed effects
  • Higher-order models: SLX, SAC/SARAR with separate lag and error weights, SDEM
  • Spatial regimes: Regime-varying coefficients with SKATER regionalization and a Chow-type LR test
  • Discrete choice: Spatial probit by linearized GMM and Albert-Chib Gibbs, ordered probit, marginal effects
  • Spatial filtering: Eigenvector spatial filtering (ESF) for Gaussian/Poisson/logit/negbin, and RE-ESF by profile REML
  • Spatial diagnostics: Moran's I (global & LISA), Geary's C, LM tests (lag, error, robust)
  • Panel diagnostics: Spatial Hausman, Chow, joint LM tests, panel Moran's I
  • Model selection: AIC/BIC, LR test, Vuong test, common factor test, Wald test
  • Weight matrices: k-NN, distance-decay, distance-band, block-diagonal, standardization, GAL/GWT I/O
  • Robust inference: Driscoll-Kraay SCC, Conley spatial HAC
  • Simulation: Six spatial DGPs, Monte Carlo experiments, power curves
  • Output: Tables.jl coefficient tables, side-by-side regression tables, LaTeX/CSV/Markdown export
  • Bayesian extension: Turing.jl integration for Bayesian spatial models
  • Visualization extension: Makie.jl plotting for LISA maps and coefficient surfaces

Installation

using Pkg
Pkg.add("SpatialEconometrics")

Quick Example

using SpatialEconometrics

# Create spatial weights
W = make_rook_weights(20)

# Fit a spatial autoregressive model
m = fit_spatial(SAR(), W, X, y)

# Diagnostics
mr = moran_global(residuals(m), W)
lm = lm_tests(y, hcat(ones(length(y)), X), W)

# Model comparison
println(model_comparison_table(m_sar, m_sdm, m_sem; names=["SAR","SDM","SEM"]))

Contents