B0872
Title: Factor-augmented regression for high dimensional time series
Authors: Dehao Dai - University of California San Diego (United States) [presenting]
Danna Zhang - University of California, San Diego (United States)
Abstract: The existing work on supervised learning of time series data often assumes that the latent factor model or time series linear regression model is the true underlying model without justifying its adequacy. To fill in such an important gap in high-dimensional inference, factor-augmented time series regression is leveraged as the alternative model to test the sufficiency of the latent factor model. The model utilizes functional dependence measures to account for a wide class of dependence structures as well as general exponential-type tails existing in factors, factor loadings, idiosyncratic errors, and regression errors. Convergence rates are provided for the estimators of components in factor models and a Gaussian approximation result is established for de-biased regularized estimators for the regression parameters. The theoretical findings are extensively validated through numerical experiments, including simulations and the analysis of real-world FRED macroeconomic data.
