Abstract:Despite the abundance of methods for variable selection and accommodating spatial structure in regression models, there is little precedent for incorporating spatial dependence in covariate inclusion probabilities for regionally varying regression models. The lone existing approach is limited by difficult computation and the requirement that the spatial dependence be represented on a lattice, making this method inappropriate for areal models with irregular structures that often arise in ecology, epidemiology, and the social sciences. Here we present a novel method for spatial variable selection in areal generalized linear models that can accommodate arbitrary spatial structures and works with a broad subset of GLM likelihoods. The method uses a latent probit model with a spatial dependence structure where the binary response is taken as a covariate inclusion indicator for area-specific GLMs. The covariate inclusion indicators arise via thresholding of latent standard normals on which we place a conditionally autoregressive prior. We propose an efficient MCMC algorithm for computation that is entirely conjugate in any model with a conditionally Gaussian representation of the likelihood, thereby encompassing logistic, probit, multinomial probit and logit, Gaussian, and negative binomial regressions through the use of existing data augmentation methods. We demonstrate superior parameter recovery and prediction in simulation studies as well as in applications to geographic voting patterns and population estimation. Though the method is very broadly applicable, we note in particular that prior to this work, spatial population estimation/capture-recapture models allowing for varying list dependence structures has not been possible.

Generalized Spatial Regression with Differential Regularization

Local Linear M‐estimation in Non‐parametric Spatial Regression

Spatial Proportional Hazards Model with Differential Regularization

Additive model building for spatial regression

Estimation and Inference for Generalized Geoadditive Models

Bayesian variable selection for spatially dependent generalized linear models

GEE-assisted Forward Regression for Spatial Latent Variable Models

Spatial confounding under infill asymptotics

Nonparametric Density Estimation for Data Scattered on Irregular Spatial Domains: A Likelihood-Based Approach Using Bivariate Penalized Spline Smoothing

Spatially Clustered Regression

Series ridge regression for spatial data on $\mathbb{R}^d$

Generalized Linear Latent Variable Models for Repeated Measures of Spatially Correlated Multivariate Data

Generalized spatial linear regression model for spatial prediction

Generalized spatial autoregressive model

A Spatially Correlated Model with Generalized Autoregressive Conditionally Heteroskedastic Structure for Counts of Crimes

Two-stage Estimators for Spatial Confounding

When Space Matters: Spatial Dependence, Diagnostics, and Regression Models

Variable Selection for Spatial Nonparametric Regression

Spatial Autologistic Model with Generalized Dependent Parameter

Local Walsh-average Regression for Spatial Autoregression Single Index Varying Coefficient Models

Spatial homogeneity pursuit of regression coefficients for large datasets