AUTOREGRESSION, OR AUTOREGRESSIVE PROCESS

(Statistics) A condition which exists whenever a lagged (i.e., prior period) value of the Dependent Variable, or the variable to be explained, appears as a regressor, that is, as an Explanatory Variable. The fundamental assumption is that future data values may be expressed as linear combinations of past observations. It is not uncommon in economics and other areas of scientific study for a variable to be influenced by its own behavior in prior periods. The problem with this equation (model) format is to insure that the lagged variable, represented below as Yt-1, is independent of the disturbance term, t. An example of a (first-order) autoregressive process, commonly termed AR(1), would be represented by: Yt = ø1Yt-1 + ð + et; where the parameter ø1 < 1, and ð is the (constant, time insensitive) trend component, and et is the residual or disturbance term associated with each observation of Yt.