Decomposition of discrete time periodically correlated and multivariate stationary symmetric stable processes (Q2572204)

Following closely the strategy of \textit{J. Rosinski} [Ann. Probab. 23, No. 3, 1163--1187 (1995; Zbl 0836.60038)] the authors investigate the spectral structure of periodically correlated (PC) symmetric \(\alpha\)-stable processes discrete in time. They find an analogous decomposition of the process into three mutually independent parts: A mixed moving average, a harmonisable process and a third component not being harmonisable. The proof starts with a decomposition of the process into two disjoint parts. One generated by a dissipative and the other by a conservative nonsingular flow. Then the dissipative part gives the mixed moving average whereas the conservative part is decomposed again into a harmonisable process and another non-harmonisable one. An analogous decomposition is proven for a stationary \(T\)-variate (\(T\in {\mathbb N}\)) symmetric \(\alpha\)-stable process \({\mathbf Y}_{n} = (Y_{n}^0, \dots , Y_{n}^{T-1})\) using that a PC symmetric \(\alpha\)-stable process can be identified with stationary \(T\)-variate symmetric \(\alpha\)-stable process.