A statistical approach using field-field correlation functions which is obtained from two-probe time domain measurement is used to characterize the radiation from complex devices. The time-frequency analysis provided by the measurement data has shown that significant emissions may only occur for a few percent of the time and a piecewise stationary model of emissions may be most appropriate. In this paper, a sorting technique is applied to short time segments of the nonstationary time-domain data provided by the measurements. These short time segments are sorted into groups according to the characteristics of the emissions, i.e. different emission processes, with the assumption that the stochastic emissions is stationary within each group (process). The ensemble of time segments associated with each group may then be used to obtain the fieldfield correlations for each process. Results of the analysis give promising insights into how to characterize complex and time dependent systems.