Unit root testing via the stationary bootstrapAAA
Date
2006Source
Journal of EconometricsVolume
133Issue
2Pages
601-638Google Scholar check
Keyword(s):
Metadata
Show full item recordAbstract
A nonparametric, residual-based stationary bootstrap procedure is proposed for unit root testing in a time series. The procedure generates a pseudoseries which mimics the original, but ensures the presence of a unit root. Unlike many others in the literature, the proposed test is valid for a wide class of weakly dependent processes and is not based on parametric assumptions on the data-generating process. Large sample theory is developed and asymptotic validity is shown via a bootstrap functional central limit theorem. The case of a least squares statistic is discussed in detail, including simulations to investigate the procedure's finite sample performance. © 2005 Elsevier B.V. All rights reserved.