6
28

De-Biasing The Lasso With Degrees-of-Freedom Adjustment

Abstract

This paper studies schemes to de-bias the Lasso in a linear model y=Xβ+ϵy=X\beta+\epsilon where the goal is to construct confidence intervals for a0Tβa_0^T\beta in a direction a0a_0, where XX has iid N(0,Σ)N(0,\Sigma) rows. We show that previously analyzed propositions to de-bias the Lasso require a modification in order to enjoy efficiency in a full range of sparsity. This modification takes the form of a degrees-of-freedom adjustment that accounts for the dimension of the model selected by Lasso. Let s0s_0 be the true sparsity. If Σ\Sigma is known and the ideal score vector proportional to XΣ1a0X\Sigma^{-1}a_0 is used, the unadjusted de-biasing schemes proposed previously enjoy efficiency if s0n2/3s_0\lll n^{2/3}. However, if s0n2/3s_0\ggg n^{2/3}, the unadjusted schemes cannot be efficient in certain a0a_0: then it is necessary to modify existing procedures by a degrees-of-freedom adjustment. This modification grants asymptotic efficiency for any a0a_0 when s0/p0s_0/p\to 0 and s0log(p/s0)/n0s_0\log(p/s_0)/n \to 0. If Σ\Sigma is unknown, efficiency is granted for general a0a_0 when \frac{s_0\log p}{n}+\min\Big\{\frac{s_\Omega\log p}{n},\frac{\|\Sigma^{-1}a_0\|_1\sqrt{\log p}}{\|\Sigma^{-1/2}a_0\|_2 \sqrt n}\Big\}+\frac{\min(s_\Omega,s_0)\log p}{\sqrt n}\to0 where sΩ=Σ1a00s_\Omega=\|\Sigma^{-1}a_0\|_0, provided that the de-biased estimate is modified with the degrees-of-freedom adjustment. The dependence in s0,sΩs_0,s_\Omega and Σ1a01\|\Sigma^{-1}a_0\|_1 is optimal. Our estimated score vector provides a novel methodology to handle dense a0a_0. Our analysis shows that the degrees-of-freedom adjustment is not needed when the initial bias in direction a0a_0 is small, which is granted under stringent conditions on Σ1\Sigma^{-1}. The main proof argument is an interpolation path similar to that typically used to derive Slepian's lemma. It yields a new \ell_\infty error bound for the Lasso which is of independent interest.

View on arXiv
Comments on this paper