A Conditional Randomization Test for Sparse Logistic Regression in High-Dimension

TL;DR

This paper introduces CRT-logit, an efficient and theoretically grounded method for variable selection in high-dimensional sparse logistic regression, improving inference accuracy and computational speed.

Contribution

It develops CRT-logit, a novel algorithm combining variable distillation and decorrelation tailored for high-dimensional logistic regression, with theoretical guarantees.

Findings

CRT-logit outperforms existing methods in simulation studies.

It achieves accurate variable selection with controlled error rates.

Demonstrates effectiveness on large-scale brain-imaging and genomics datasets.

Abstract

Identifying the relevant variables for a classification model with correct confidence levels is a central but difficult task in high-dimension. Despite the core role of sparse logistic regression in statistics and machine learning, it still lacks a good solution for accurate inference in the regime where the number of features $p$ is as large as or larger than the number of samples $n$. Here, we tackle this problem by improving the Conditional Randomization Test (CRT). The original CRT algorithm shows promise as a way to output p-values while making few assumptions on the distribution of the test statistics. As it comes with a prohibitive computational cost even in mildly high-dimensional problems, faster solutions based on distillation have been proposed. Yet, they rely on unrealistic hypotheses and result in low-power solutions. To improve this, we propose \emph{CRT-logit}, an algorithm that combines a variable-distillation step and a decorrelation step that takes into account the geometry of $\ell_1$-penalized logistic regression problem. We provide a theoretical analysis of this procedure, and demonstrate its effectiveness on simulations, along with experiments on large-scale brain-imaging and genomics datasets.