Phase disturbance due to atmospheric turbulence aects fringe tracking. Algorithms aimed at fringe parameter identification are based on interferometric models that have to be carefully adapted to the interfered beams, including sources of variability. All the information is contained in the collected signals, and how to extract it is a major research problem. This work aims to determine some stochastic properties of the signals from the data that are useful for modelling. We apply statistical techniques to real interferometric signals. We examine the composition of signals before and after the combination using time series analysis, and we separate pure random eects and peculiar features from trends. Multivariate regression analysis allows us to isolate noise components due to the interference physical process.