Application of the covariance analysis by the “sliding window” method to assess the relationship of non-stationary time series

A covariance analysis method is described to solve the problem of estimating the dynamics of the relationship between two non-stationary time series, represented by behavioral and/or physiological data. The application of the “sliding window” method is proposed for the possibility of analyzing the degree of connectivity of time series at different epochs. The covariance analysis, unlike traditionally used methods, allows to take into account the magnitude of the dynamics of time series indicators when evaluating the relationship. When comparing covariance and correlation analyses, a high degree of stability of covariance analysis to signal noise was noted.