83
40
v1v2v3 (latest)

Notes on the dimension dependence in high-dimensional central limit theorems for hyperrectangles

Abstract

Let X1,,XnX_1,\dots,X_n be independent centered random vectors in Rd\mathbb{R}^d. This paper shows that, even when dd may grow with nn, the probability P(n1/2i=1nXiA)P(n^{-1/2}\sum_{i=1}^nX_i\in A) can be approximated by its Gaussian analog uniformly in hyperrectangles AA in Rd\mathbb{R}^d as nn\to\infty under appropriate moment assumptions, as long as (logd)5/n0(\log d)^5/n\to0. This improves a result of Chernozhukov, Chetverikov & Kato [Ann. Probab. 45 (2017) 2309-2353] in terms of the dimension growth condition. When n1/2i=1nXin^{-1/2}\sum_{i=1}^nX_i has a common factor across the components, this condition can be further improved to (logd)3/n0(\log d)^3/n\to0. The corresponding bootstrap approximation results are also developed. These results serve as a theoretical foundation of simultaneous inference for high-dimensional models.

View on arXiv
Comments on this paper