A CENTRAL LIMIT THEOREM OF FOURIER TRANSFORMS OF STRONGLY DEPENDENT STATIONARY PROCESSES

From MaRDI portal





Let \(x_ t(t=0,1,...,T-1)\) be the observations from a strictly stationary process and h(t), \(t\in [0,1]\), be a real-valued function of bounded variation. Define the finite Fourier transform by \[ d_ T(\lambda)=\sum^{T-1}_{i=0}h(t/T)x_ t \exp (-i\lambda t). \] The finite Fourier transform is an important entity for analysing the structure of the process. Its asymptotic properties have been studied by many workers under the assumption that autocorrelations and higher-order cumulants of the processes are absolutely summable, which implies that observations well separated in time are weakly dependent. The author proves that, if the process is strictly stationary and satisfies some conditions on its higher-order cumulants, the finite Fourier transforms at different frequencies are asymptotically independent and normally distributed. The assumptions proposed by the author are satisfied by many strongly dependent stationary processes: e.g. fractional Gaussian noise models and fractional autoregressive integrated moving-average (ARIMA) models.




Cited in
(24)








This page was built for publication: A CENTRAL LIMIT THEOREM OF FOURIER TRANSFORMS OF STRONGLY DEPENDENT STATIONARY PROCESSES

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3034606)