High dimension complex dynamical systems, such as those found in physiological processes, produce time series which are often accompanied by nonstationarity. In many cases, the nonstationarity is caused by a physiologically significant event such as the prelude to ventricular fibrillation in cardiac arrest or the change of stasis by introducing pharmaceuticals. A need exists to be able to detect and monitor this change. Most conventional attempts at addressing this problem involve segmenting the time series and evaluating the statistics of the segments. The difficulty with this approach is that the nature of the nonstationarity can be transient, such that it is bounded by two, or more, regions of stationarity. This oscillation may continue for a significant portion of the time series. This research will address the underlying statistical justification for asserting stationarity and linearity as well as the use of segmentation time series analysis techniques.
