NOISE IN AUTOREGRESSIVE TIME-SERIES

In this paper a method is devised to determine the size of external shocks (noise) in an autoregressive timeseries, when nothing is known about the lag structure, not even its order. The technique considers the timeseries to be a hyperbody embedded in a high dimensional space, and makes use of the "correlation integral", as introduced in the physical sciences by Grassberger and Procaccia, to measure this body's thickness. Uniformly distributed noise is considered at first, after which the technique is adapted to normal and other distributions. © 1991.