We propose a new method, the "Auto-SLEX" method, for analyzing bivariate non-stationary processes. The Auto-SLEX method is a procedure that automatically segments the time series into approximatively stationary blocks and automatically estimates the time-varying spectra and coherence.

A consistent estimator for the spectral density of a stationary random process can be obtained by smoothing the periodigrams across frequency. An important component of smoothing is the choice of the span. In this paper, we propose a span selector originally developed for use in fitting...