Infinite Order Autoregressive Representations of Multivariate Stationary Stochastic Processes.
Abstract
While it is well-known that every purely nondeterministic full rank q- variate weakly stationary stochastic process (X sub n) with the spectral density W has an (infinite order) one-sided moving average representation, not every such process can have a mean convergent (infinite order) autoregressive representation (ARR) and the problem of ARR of such processes has not received the attention which it deserves. Due to the importance of ARR in prediction theory, and particularly in the statistical theory of multivariate time series, this paper is devoted to the problem of finding the weakest condition on W which guarantees the existence of an infinite order ARR for (X sub n).
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1984
- Accession Number
- ADA147863
Entities
People
- M. Pourahmadi
Organizations
- University of North Carolina at Chapel Hill