Motivated by statistical inference problems in high-dimensional time series data analysis, we first derive non-asymptotic error bounds for Gaussian approximations of sums of high-dimensional dependent random vectors on hyper-rectangles, simple convex sets and sparsely convex sets. We investigate the quantitative effect of temporal dependence on the rates of convergence to a Gaussian random vector over three different dependency frameworks ($\alpha$-mixing, $m$-dependent, and physical dependence measure). In particular, we establish new error bounds under the $\alpha$-mixing framework and derive faster rate over existing results under the physical dependence measure. To implement the proposed results in practical statistical inference problems, we also derive a data-driven parametric bootstrap procedure based on a kernel estimator for the long-run covariance matrices. We apply the unified Gaussian and bootstrap approximation results to test mean vectors with combined $\ell^2$ and $\ell^\infty$ type statistics, change point detection, and construction of confidence regions for covariance and precision matrices, all for time series data.
翻译:针对高维时间序列数据分析中的统计推断问题,本文首先推导了超长方体、简单凸集和稀疏凸集上高维相关随机向量之和的高斯逼近的非渐近误差界。我们针对三个不同的依赖框架 ($\alpha$-混合,$m$-相关和物理相关度度量) 研究了时间依赖性对收敛速率到高斯随机向量的影响。特别地,我们在 $\alpha$-混合框架下建立了新的误差界,并在物理相关度量下比现有结果提高了收敛速率。为了在实际统计推断问题中实现所提出的结果,我们还基于长期协方差矩阵的核估计器导出了数据驱动的参数引导程序。我们将统一的高斯和引导逼近结果应用于包括合并 $\ell^2$ 和 $\ell^\infty$ 类型统计、变点检测和协方差和精度矩阵置信区间构建的时间序列数据上。