R implementation of a spatially-extended ridge regression using the EM algorithm
This repository provides an R implementation of the Expectation-Maximization (EM) algorithm for Generalized Ridge Regression with spatial covariates.
The method is evaluated via a simulation study comparing regression coefficient estimates under different spatial priors:
- Matérn covariance
- Conditional Autoregressive (CAR) model
- Standard (non-spatial) Ridge
The provided simulation illustrates:
- Generating spatially correlated covariates and responses using a Matérn process
- Running the EM algorithm for each covariance structure
- Comparing the estimated β coefficients to the true underlying values
The full approach is described in:
Obakrim, S. et al. “EM algorithm for generalized Ridge regression with spatial covariates,” Environmetrics, 2024. doi:10.1017/eds.2022.35