Supplementary Material for "A simple approach for local and global variable importance in nonlinear regression models"

This repo contain supplementary tables from the manuscript entitled: "A simple approach for local and global variable importance in nonlinear regression models". Here, we provide genome-wide results for all single nucleotide polymorphisms (SNPs) in a heterogenous stock of mice data set from the Wellcome Trust Centre of Human Genetics while analyzing, body weight, high-density lipoprotein (HDL), and percentage of CD8+ cells. Listed are variable importance values using the “RelATive cEntrality” (RATE) and the “GlObal And Local Score" (GOALS) approaches for each SNP as computed via Gaussian process regression. As a direct comparison, we also include variable importance scores for each SNP after running a random forest (RF), a gradient boosting machine (GBM), and a Bayesian additive regression tree (BART).