Bootstrap Approximating the Mean of Long Memory Time Series

Abstract

This paper evaluated the finite sample performances of three bootstrap methods, sieve AR bootstrap (SARB), fractional differencing sieve bootstrap (FDSB) and fractional differencing block bootstrap (FDBB), in approximating the mean of long memory time series. Extensive simulations show that the FDBB method has more stable approximate results in most cases and can more accurately approximate the mean distribution of long memory time series than the other two methods.