To generate multivariate synthetic sequences that will resemble multivariate historic sequences in terms of means, standard deviations, skewnesses, lag‐one serial correlation, and lag‐zero cross‐correlations, a multivariate, weakly stationary process is used. Techniques of regionalization and maximum likelihood estimates may be used to minimize the biases associated with the estimates of the parameters that characterize the historic sequences.
N. C. Matalas (Fri,) studied this question.