A Note on Nonlocal Prior Method

TL;DR

This paper introduces a new nonlocal prior for high-dimensional variable selection, enhancing robustness to hyperparameters and improving detection of small signals.

Contribution

It proposes a novel nonlocal prior that is robust to hyperparameter choices and capable of detecting weaker signals in high-dimensional data.

Findings

Improved variable selection performance in high dimensions

Robustness to hyperparameter variations

Enhanced detection of small signals

Abstract

We propose a new class of nonlocal prior to improve the performance of variable selection in high dimensional setting. We prove our new prior possesses the robustness to hyper parameter settings and is able to detect smaller decreasing signals.