Nonregular and Minimax Estimation of Individualized Thresholds in High Dimension with Binary Responses

DOI10.1214/22-AOS2188arXiv1905.10888MaRDI QIDQ6319397FDOQ6319397

Authors: Huijie Feng, Yang Ning, Jiwei Zhao 

Publication date: 26 May 2019

Abstract: Given a large number of covariates $Z$, we consider the estimation of a high-dimensional parameter $heta$ in an individualized linear threshold $het{a}^{T}Z$ for a continuous variable $X$, which minimizes the disagreement between $extsign(X-het{a}^{T}Z)$ and a binary response $Y$. While the problem can be formulated into the M-estimation framework, minimizing the corresponding empirical risk function is computationally intractable due to discontinuity of the sign function. Moreover, estimating $heta$ even in the fixed-dimensional setting is known as a nonregular problem leading to nonstandard asymptotic theory. To tackle the computational and theoretical challenges in the estimation of the high-dimensional parameter $heta$, we propose an empirical risk minimization approach based on a regularized smoothed loss function. The statistical and computational trade-off of the algorithm is investigated. Statistically, we show that the finite sample error bound for estimating $heta$ in $el{l}_2$ norm is $$, where $d$ is the dimension of $heta$, $s$ is the sparsity level, $n$ is the sample size and $$ is the smoothness of the conditional density of $X$ given the response $Y$ and the covariates $Z$. The convergence rate is nonstandard and slower than that in the classical Lasso problems. Furthermore, we prove that the resulting estimator is minimax rate optimal up to a logarithmic factor. The Lepski's method is developed to achieve the adaption to the unknown sparsity $s$ and smoothness $$. Computationally, an efficient path-following algorithm is proposed to compute the solution path. We show that this algorithm achieves geometric rate of convergence for computing the whole path. Finally, we evaluate the finite sample performance of the proposed estimator in simulation studies and a real data analysis.

Mathematics Subject Classification ID

Density estimation (62G07) Asymptotic properties of nonparametric inference (62G20) Classification and discrimination; cluster analysis (statistical aspects) (62H30) Minimax procedures in statistical decision theory (62C20)

This page was built for publication: Nonregular and Minimax Estimation of Individualized Thresholds in High Dimension with Binary Responses

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6319397)