Stature is a key concept in the study of human past since along with the biological information it provides overall trends about the standards of living. However, stature data on historic and prehistoric populations are still limited, especially for key temporospatial settings of antiquity such as ancient Greece. We collected osteometric data from 35 male and 33 female individuals (n = 68) from a Roman Period necropolis (146 BC-330 AD) in northern Greece and we applied the anatomical stature estimation method (Raxter et al. 2006). We compared this estimation with the results of 20 regression equations (both ordinary least squares and reduced major axis methods) and we examined the factors that affect their accuracy such as body proportions and the chronology and geography of reference series. For our analyses, we calculated the percent prediction error (%PE) produced by each regression equation for males and females separately. We introduced the total %PE to evaluate each equation's effectiveness on both sexes simultaneously. We calculated long bone ratios and the Euclidean distance between the Greek dataset and the reference series. According to the anatomical method, males from Northern Greece had a mean stature value of 168.2 cm ± 5.38 and females of 156.9 cm ± 5.27. The regression equations of Vercellotti et al. (2009) provided the best estimations in both sexes. Estimation errors (%PE) were not statistically different between the ordinary least squares and reduced major axis equations. The chronological affinity in body proportions between the target population and the reference series could provide significant evidence for the prediction of the optimal regression formulae. With this paper we provide the first osteometric dataset for the anatomical stature estimation method from ancient Greece and we suggest the most suitable regression equations for this key region of the ancient World.