Analysis of stellar variability based on polynomial fittings of its light curve

A method for analysis of stellar light curves (LCs), and in principle, time series, is presented here. It is based on polynomial fitting of the LC with increasing polynomial degree M. Two basic parameters from each polynomial fit of are regarded: residual mean square deviation (SD) of the data in respect to the polynomial fit, SM, and the half of the absolute average deviation of the polynomial (PD) in respect to the average of the data, PM. In each cases the maximal regarded polynomial degree L corresponds to the minimal SD, SL. (In practice the polynomial with degree L + 1 produces larger standard deviation because the number of data is not large and the calculation errors accumulate.) We found well pronounced anti-correlations between SM and PM when M changes from 1 to L. The respective slope PD/SD turns out to be a useful quantify parameter of the LC, characterizing the presence of significant coarse details and giving possibilities for classification of LCs.