Data acquired by GOLF experiment onboard SOHO during ~20 months has been analysed. GOLF is a disk-integrated sunlight experiment therefore biased to observe very low ell modes (ell < 3). However the excellent ratio signal to background (S/B) achieved permits also the detection of some modes with 4 < ell < 6 and those from n >= 8 up to the cut-off frequency and even beyond with the pseudomodes signal. Due to the nature of the excitation function of the modes (which we will call noise) the line shapes look "spiky" rising the possibility (the necessity?) to use several methods of analysis of such data: a) FFT spectrum and maximum likelihood fitting technique, b) averaged FFT spectra of subseries and least squares fitting technique, c) HD spectra and least squares fit. Moreover, as a consequence of the characteristics of the spectrum: 1 <= S/B <= 104, linewidths of 0.1 <= Γ <= 40 muHz and the presence of noise, several fit strategies have to be defined in order to fit appropriately and accurately the mode line shapes. In this poster these methods and fit strategies will be explained and comparative results will be presented.