Time series in m dimensions: Autoregressive models

Abstract
Spatially dependent autoregressive models in m dimensions are defined. The conditions for stationarity and invertibility are determined. The autocorrelation function and Yule-Walker equations are obtained for the general case, and as particular cases for special discrete values for various grids in plane and for orders 1 and 2 in time. The spectra are obtained for these particular cases, and some results for the partial autocorrelation function. All results are new. The notation, definitions, and assumptions are those given by Voss et al. (1980). We assume stationarity of z over time t, where an m dimensional vec 12m tor. We assume the covariance structure as given by Hannan (1970), 2 with and all covariances existing. Nonstationary models will be considered in later papers.

This publication has 4 references indexed in Scilit: