This contribution discusses the estimation of an invertible functional time series through fitting of functional moving average processes. The method uses a functional version of the innovations algorithm and dimension reduction onto a number of principal directions. Several methods are suggested to automate the procedures. Empirical evidence is presented in the form of simulations and an application to traffic data.
Seitenangaben Beitrag:
51-58
Stichworte:
Order Selection, Functional Principal Component Analysis, Functional Principal Component, Principal Subspace, Innovation Algorithm