This repository contains scripts, files, and figures related to the simulation study and the real data application described in the paper "Regularized Principal Spline Functions to Mitigate Spatial Confounding" by Carlo Zaccardi, Pasquale Valentini, Luigi Ippoliti and Alexandra M. Schmidt (arXiv link).
source("install_packages.r")
-
simulations.r is the main script for running all the simulations using the different approaches compared in the paper.
-
folder "func" contains the main functions for the simulations:
- data_generate.r contains the function to generate the data for the simulations study.
- setting.xslx contains data table with parameters used to generate data. Each row correspond to a configuration.
- sim_functions.r contains the functions to do the simulations using our approach.
- sim_spectralAdj.r contains functions to run simulations using the Spectral Adjustment approach by Guan et al. (2023).
- misc.r contains auxiliary functions.
-
folder "results" contains some R files that are produced by the simulations. These have been used to create the figures.
-
folder "plots" contains all the figures in the paper.
-
plot results.r produces figures related to the simulation study.
-
plot results_suppMat.r produces figures in the Supplementary Web Material.
-
ozone.r reproduces the real data application.
-
SSconf_1.0.tar.gz is a package containing the function
spsl()to implement our approach. This is written in C++ using the Armadillo library, and is embedded in an R package to allow for parallelization.
The code has been tested on Windows 11 and Linux (Ubuntu) machines.
The numbers of iterations of the MCMC algorithms are small for the purpose of illustration. Increase their values for morre accurate results.