On empirical multiple time series analysis
Emanuel Parzen · Project Euclid (Cornell University) · 1967
Abstract : It is natural to think of the theory of time series analysis as composed of two parts, foundations (a probabilistic part involving deep mathematics and based on the unrealistic assumption that one knows the probability law of the time series) and empirical (in which one considers statistical and computational procedures). This paper outlines the problem of empirical multiple time series analysis, indicates some theoretical questions which require further investigation, presents a unified but heuristic exposition of crossspectral analysis, and states some new results on the asymptotic sampling theory of partial cross-spectra. (Author)