Abstract:This paper considers point and interval estimation of the l(q) loss of an estimator in high-dimensional linear regression with random design. We establish the minimax rate for estimating the l(q) loss and the minimax expected length of confidence intervals for the l(q) loss of rate-optimal estimators of the regression vector, including commonly used estimators such as Lasso, scaled Lasso, square-root Lasso and Dantzig Selector. Adaptivity of confidence intervals for the l(q) loss is also studied. Both the setting of the known identity design covariance matrix and known noise level and the setting of unknown design covariance matrix and unknown noise level are studied. The results reveal interesting and significant differences between estimating the l(2) loss and l(q) loss with 1 <= q < 2 as well as between the two settings. New technical tools are developed to establish rate sharp lower bounds for the minimax estimation error and the expected length of minimax and adaptive confidence intervals for the l(q) loss. A significant difference between loss estimation and the traditional parameter estimation is that for loss estimation the constraint is on the performance of the estimator of the regression vector, but the lower bounds are on the difficulty of estimating its l(q) loss. The technical tools developed in this paper can also be of independent interest.

Minimax Estimator of Regression Coefficient in Normal Distribution under Balanced Loss Function

Linearly Admissible Estimators of Stochastic Regression Coefficient under Balanced Loss Function

The Superiority of the Bayes Linear Unbiased Minimum Variance Estimator under Balanced Loss Function

Reliability Analysis of Cox Models under the Balanced Loss Function: Based on Records

Quasi-minimax Estimation in the General Linear Regression Model

Linearly Admissible Estimators of the Common Mean Parameter under Balanced Loss Function

Non-parametric Shrinkage Mean Estimation for Quadratic Loss Functions with Unknown Covariance Matrices

Sparse Estimation Via ℓ_q Optimization Method in High-Dimensional Linear Regression

Bayes Estimation for the Half-normal Distribution under Different Loss Function

A Non-Parametric Shrinkage Mean Estimation for Arbitrary Quadratic Loss Functions and Unknown Covariance Matrices

Sparse Estimation Via Lower-Order Penalty Optimization Methods in High-Dimensional Linear Regression.

Nonnegative Estimation and Variable Selection under Minimax Concave Penalty for Sparse High-Dimensional Linear Regression Models

A New Recursive Bounded GM Estimator for Regression

Estimator of the Scale-Parameter in a Class of the Exponential Family under Weighted Quadratic Loss Function

Bayesian and minimax estimators of loss

Accuracy Assessment for High-dimensional Linear Regression

Polynomials shrinkage estimators of a multivariate normal mean

Asymptotic Normality and Confidence Intervals for Prediction Risk of the Min-Norm Least Squares Estimator.

Non-minimaxity of debiased shrinkage estimators

Minimax rates of l p-losses for high-dimensional linear regression models with additive measurement errors over l q-balls

Optimal convex $M$-estimation via score matching