CHANGE DETECTION MODEL FOR SERIALLY CORRELATED DATA

Develops a time series model for detecting change in psychophysiological data from a single S following stimulation. A stationary model is assumed for the periods before stimulation; it is assumed that the S returns to a steady or basal state between stimuli. A 1st-order autoregression (Markov process) is fit to all prestimulus data for a single S. Predictions are then made into the poststimulus regions. If the time series is Gaussian and if there is no response to the stimulus, the differences between 1-step predictions and the corresponding observations are independent and normally distributed with mean zero and constant variance which can be estimated. Statistical tests for change are constructed to determine whether the differences are significant. The objective is to detect any deviations from the stationary structure of the prestimulus periods. (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1969 American Psychological Association.