One of the goals of SSA in time-series analysis is to identify a signal or signals in an observed record against a background of noise (e.g., Rasmusson et al., 1990; Plaut et al., 1995). If we are lacking reasons to anticipate a particular regularity based on physical theory, then we need to decide at what confidence level we can reject the notion that features identified in the record have occurred by chance. As suggested by Allen (1992) we can think of “occurring by chance” to mean “attributable to the stochastic component of the record.” So it is necessary to calculate the probability of these stochastic features exceeding certain levels. It is then possible to estimate the chance of a given feature being the product of random fluctuations or the product of some physical phenomenon.
KeywordsMonte Carlo Signal Detection Noise Process Monte Carlo Approach Singular Spectrum Analysis
Unable to display preview. Download preview PDF.