Virgin olive oil samples of eight varieties from four regions (North Aegean, South Aegean, Mediterranean, and Southeastern) of Turkey were discriminated using fatty acid and sterol composition. Principle component analysis represented a separation of South Aegean olive oils from the rest of the sample groups, that mainly depend on Stigmasterol, β-sitosterol, Δ5,24-Stigmastadienol, Δ7-Avenasterol, C17:0, and C17:1 variables. Except few overlaps, North Aegean samples were also discriminated with Mediterranean and Southeastern samples. The varietal separation was not interpretable by itself but since all samples from South Aegean region were Memecik variety, regional separation has become clearer. Soft independent modeling of class analogy shows good separation between North and South Aegean samples with only a few exceptions. The number of misestimated samples was high at Mediterranean and Southeastern models on Coomans’ plots because of high variance within each group. Partial least squares discrimination analysis was more successful than Soft independent modeling of class analogy. The prediction capabilities of South Aegean and North Aegean models were better than others. Root mean squared error of prediction and goodness of prediction were 0.092 and 0.961 for South Aegean, 0.182, and 0.853 for North Aegean, respectively. Unlikely to soft independent modeling of class analogy, Southeastern and Mediterranean samples were not rejected but remained as “uncertain” on partial least squares discrimination analysis with the help of its algorithm.