Support estimation in high-dimensional heteroscedastic mean regression

A current strand of research in high-dimensional statistics deals with robustifying the available methodology with respect to deviations from the pervasive light-tail assumptions. In this paper we consider a linear mean regression model with random design and potentially heteroscedastic, heavy-tailed errors, and investigate support estimation in this framework. We use a strictly convex, smooth variant of the Huber loss function with tuning parameter depending on the parameters of the problem, as well as the adaptive LASSO penalty for computational efficiency. For the resulting estimator we show sign-consistency and optimal rates of convergence in the $\ell_\infty$ norm as in the homoscedastic, light-tailed setting. In our analysis, we have to deal with the issue that the support of the target parameter in the linear mean regression model and its robustified version may differ substantially even for small values of the tuning parameter of the Huber loss function. Simulations illustrate the favorable numerical performance of the proposed methodology.

• Publication year

  2020

• Venue

  arXiv: Statistics Theory

• Publication date

  2020-11-03

• Fields of study

  Mathematics

• Identifiers
  arXiv 2011.01591
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• High-Dimensional Probability: An Introduction with Applications in Data Science

  2020influential reference
• Estimating the Lasso's Effective Noise

  2020cited by this paper
• Nonlinear Optimization

  2019influential reference
• Optimal Variable Selection and Adaptive Noisy Compressed Sensing

  2018cited by this paper
• High-Dimensional Probability

  2018cited by this paper
• Adaptive Huber Regression

  2017influential reference
• Estimation of high dimensional mean regression in the absence of symmetry and light tail assumptions

  2017influential reference
• High-dimensional generalizations of asymmetric least squares regression and their applications

  2016cited by this paper
• I-LAMM FOR SPARSE LEARNING: SIMULTANEOUS CONTROL OF ALGORITHMIC COMPLEXITY AND STATISTICAL ERROR.

  2015cited by this paper
• Semismooth Newton Coordinate Descent Algorithm for Elastic-Net Penalized Huber Loss Regression and Quantile Regression

  2015influential reference
• Statistical Learning with Sparsity: The Lasso and Generalizations

  2015cited by this paper
• Support recovery without incoherence: A case for nonconvex regularization

  2014influential reference
• Introduction to High-Dimensional Statistics

  2014cited by this paper
• The Lasso Problem and Uniqueness

  2012influential reference
• ADAPTIVE ROBUST VARIABLE SELECTION.

  2012influential reference
• The adaptive and the thresholded Lasso for potentially misspecified models (and a lower bound for the Lasso)

  2011cited by this paper
• ℓ[subscript 1]-penalized quantile regression in high-dimensional sparse models

  2011cited by this paper
• Challenging the empirical mean and empirical variance: a deviation study

  2010influential reference
• L1-Penalised quantile regression in high-dimensional sparse models

  2009cited by this paper
• Adaptive Lasso for High Dimensional Regression and Gaussian Graphical Modeling

  2009influential reference
• Sharp Thresholds for High-Dimensional and Noisy Sparsity Recovery Using $\ell _{1}$ -Constrained Quadratic Programming (Lasso)

  2009influential reference
• ANALYZING THE RANDOM COEFFICIENT MODEL NONPARAMETRICALLY

  2009cited by this paper
• Information-Theoretic Limits on Sparsity Recovery in the High-Dimensional and Noisy Setting

  2007influential reference
• Concentration inequalities and model selection

  2007cited by this paper
• The Adaptive Lasso and Its Oracle Properties

  2006cited by this paper
• On Model Selection Consistency of Lasso

  2006cited by this paper
• Distributions generated by perturbation of symmetry with emphasis on a multivariate skew t‐distribution

  2003influential reference
• Variable Selection via Nonconcave Penalized Likelihood and its Oracle Properties

  2001cited by this paper
• Two deterministic half-quadratic regularization algorithms for computed imaging

  1994cited by this paper
• Asymmetric Least Squares Estimation and Testing

  1987cited by this paper
• Robust Estimation of a Location Parameter

  1964cited by this paper

• No citing papers are available for this paper.