Quantile inference for nonstationary processes with infinite variance innovations

Abstract Based on the quantile regression, we extend Koenker and Xiao (2004) Ling McAleer (2004)’s works from finite-variance innovations to infinite-variance innovations. A robust t -ratio statistic test for unit-root a re-sampling method approximate critical values of are proposed in this paper. It is shown that limit distribution functional stable processes Brownian bridge. The finite sample studies show always performs significantly better than conventional tests based least squares procedure, such as Augmented Dick Fuller (ADF) Philliphs-Perron (PP) test, sense power size when disturbances exist. Also, Kolmogorov-Smirnov (QKS) Cramer-von Mises (QCM) considered, but they perform poor size, respectively. An application Consumer Price Index nine countries also presented.