A0626
Title: Uniform inference for two step quantile regression process
Authors: Jayeeta Bhattacharya - University of Southampton (United Kingdom) [presenting]
Abstract: In this paper, we consider quantile regression (QR) process which involves use of estimated parameters, obtained from a preliminary first stage of estimation. The general class of the two-step QR process finds wide applicability, and some new motivating examples like estimation of factor models, are discussed. Under simple restrictions on the convergence rate of the first stage and some other regularity conditions, we establish the Bahadur representation of the two-stage QR estimator that holds uniformly across quantile levels. The result is used to establish the functional central limit theorem and obtain uniform confidence bands for the QR slope parameters. An application based on macroeconomic forecasting for analysing growth-at-risk illustrates the applicability of the uniform analysis for the two-step QR process.
