Modeling non-stationary long-memory signals with large amounts of data
Abstract
We consider the problem of modeling long-memory signals using piecewise fractional autoregressive integrated moving average processes. The signals considered here can be segmented into stationary regimes separated by occasional structural break points. The number as well as the locations of the break points and the parameters of each regime are assumed to be unknown. An efficient estimation method which can manage large amounts of data is proposed. This method uses information criteria to select the number of structural breaks. Its effectiveness is illustrated by Monte Carlo simulations.