1995A&AS..113..387B

A new approach to the frequency analysis of time series is presented. It consists in a Vanicek-like extension of Stellingwrf's method. Our MPDM (Multi-Phase Dispersion Minimization) procedure estimates the scatter of the data about a derived n-dimensional hypersurface, where n-1 co-ordinates are used to phase the measurements with known frequencies and/or to preserve the time information in order to represent possible long term variations, the last one is the phase referred to a trial frequency. Tests performed using both synthetic data and few interesting series of photometric or spectroscopic observations show the validity of our method in its definite ambit. It proves competitive dealing with pseudoperiodic time series (i.e. series which approximate periodic functions of time only inside small time ranges), nonlinear combinations of two periodicities and signals containing long term trends and/or more than one frequency with at least one strongly non-sinusoidal component. These results make the MPDM approach complementary to the existing ones.