近期许多论文研究高维平稳高斯向量自回归（VAR）模型中的估计和推理。然而，在经济学和金融学的许多应用中，相关随机变量可能具有厚尾分布，可能非线性相关，并且可能仅取正值，这使得高斯VAR模型的现有方法不适用。为了适应这些特征，本论文做了一种高维平稳高斯copula VAR模型，并运用一种简单方法来估计和推断依赖高斯copula VAR过程的转换矩阵。本文基于大方差的秩估计和变换的潜在高维高斯过程的自协方差矩阵进行估计，并根据估计量的收敛速率，开发了格兰杰因果关系的去偏推断。
Assistant Professor Hyeonseok Park's Paper Accepted for Publication in the Journal of Econometrics
Oct. 9, 2023
Hyeonseok Park, IAER Assistant Professor, had his paper accepted for publication in the Journal of Econometrics on July 27, 2023. Entitled "Estimation and inference in a high-dimensional semiparametric Gaussian copula vector autoregressive model", the paper was co-authored with Prof. Yanqin Fan and Prof. Fang Han at the University of Washington.
Estimation and inference in high-dimensional stationary Gaussian vector autoregressive (VAR) models have been considered in recent works. In many applications in economics and finance, however, the random variables of interest may have fat-tailed distributions, may be nonlinearly related, and may take positive values only, rendering existing methods for the Gaussian VAR model inapplicable. To accommodate these features, this paper proposes a high-dimensional stationary Gaussian copula VAR model and develops simple methods for the estimation and inference of the transition matrices characterizing dependence of the Gaussian copula VAR process. Our estimator is based on rank estimators of the large variance and auto-covariance matrices of a transformed latent high-dimensional Gaussian process. We derive rates of convergence of our estimator based on which we develop de-biased inference for Granger causality.