The main aim of this study is to obtain site terms from single-station sigma analysis and to compare them with the site functions resulting from different techniques. The dataset consists of 1764 records from 322 micro- and moderate-size local earthquakes recorded by 29 stations in western Turkey. Median models were derived from S-wave Fourier amplitude spectra for selected 22 frequencies, by utilizing the MLR procedure which performs the maximum likelihood (ML) estimation of mixed models where the fixed effects are treated as random (R) effects with infinite variance. At this stage, b (geometrical spreading coefficient) and Q (quality factor) values were decomposed, simultaneously. The residuals of the median models were examined by utilizing the single-station sigma analysis to obtain the site terms of 29 stations. Sigma for the median models is about 0.422 log10 units and decreases to about 0.308, when the site terms from the single-station sigma analysis were considered (27% reduction). The event-corrected within-event standard deviations for each frequency are rather stable, in the range 0.19–0.23 log10 units with an average value of 0.20 (± 0.01). The site terms from single-station sigma analysis were compared with the site function estimates from the horizontal-to-vertical-spectral-ratio (HVSR) and generalized inversion (INV) techniques by Akyol et al. (2013) and Kurtulmus and Akyol (2015), respectively. Consistency was observed between the single-station sigma site terms and the INV site transfer functions. The results imply that the single-station sigma analysis could separate the site terms with respect to the median models.