ABSTRACT
The two types of situations are roughly speaking the time-series analogues of correlation and regression. In order to describe a bivariate process it is useful to know the first- and second-order moments. For a univariate process, the first-order moment is the mean while the second-order moment is the autocovariance function, which includes the variance as a special case at lag zero. The ‘obvious’ way of estimating the cross-covariance and cross-correlation functions is by means of the corresponding sample functions. Cross-spectral analysis is a technique for examining the relationship between two series in the frequency domain. The technique may be used for two time series that ‘arise on a similar footing’ and then the coherency spectrum is perhaps the most useful function. It measures the linear correlation between two series at each frequency and is analogous to the square of the ordinary product-moment correlation coefficient.
