The Complete Lasso Tradeoff Diagram

TL;DR

This paper introduces a comprehensive tradeoff diagram for the Lasso in high-dimensional regression, illustrating the asymptotic relationship between false discovery rate and power across all achievable pairs, regardless of signal strength.

Contribution

It provides the first complete tradeoff diagram for Lasso, identifying all asymptotically attainable FDR-power pairs under linear sparsity and random designs, improving previous diagrams.

Findings

The diagram characterizes all achievable FDR-power pairs.

It reveals fundamental constraints on FDR and power tradeoffs.

Simulation studies confirm the diagram's accuracy.

Abstract

A fundamental problem in the high-dimensional regression is to understand the tradeoff between type I and type II errors or, equivalently, false discovery rate (FDR) and power in variable selection. To address this important problem, we offer the first complete tradeoff diagram that distinguishes all pairs of FDR and power that can be asymptotically realized by the Lasso with some choice of its penalty parameter from the remaining pairs, in a regime of linear sparsity under random designs. The tradeoff between the FDR and power characterized by our diagram holds no matter how strong the signals are. In particular, our results improve on the earlier Lasso tradeoff diagram of arXiv:1511.01957 by recognizing two simple but fundamental constraints on the pairs of FDR and power. The improvement is more substantial when the regression problem is above the Donoho--Tanner phase transition. Finally, we present extensive simulation studies to confirm the sharpness of the complete Lasso tradeoff diagram.