Time: Thursday, December 15th 10:00-11:30

Venue: Tencent Conference 544-203-135

Topic: On the Analysis of Quantile Forward Regression

Speaker Introduction: CHEN Hongqi , PhD in Economics, University of Illinois. Research direction: Econometrics.

Organizer: Bay Area International Business School, Beijing Normal University

Report Content:

This paper investigates the theoretical properties of Quantile Forward Regression methods under a high-dimensional linear quantile model. Two quantile forward selection regression methods are considered: the K-step and t-threshold.This paper proves the prediction bounds in the non-asymptotic case for both methods, and proves the asymptotic convergence properties of their estimated parameters and asymptotic approximation properties to the optimal prediction loss for the K-step method. Through Monte Carlo simulations, this paper compares the limited-sample performance of the quantile forward selection regression method with the common quantile-penalized regression method. The results show that the quantile forward selection regression method has certain advantages in the minimization of prediction loss and the selection of real variables. The thesis concludes by applying the quantile forward selection regression method to two macroeconomic applications in the case of high-dimensional data: the prediction of macro-growth risk and the test of the theory of international growth convergence.