Age estimation of skeletal remains is often biased due to differences between the age structure of the target sample and that of the reference population on which a particular method is based. This study analyzes a documented series of 130 skulls using the Meindl-Lovejoy method to compare of several age estimation algorithms trained on reference samples with varying demographic profiles. The results demonstrate that the estimated mean age of the test series systematically depends on the mean age of individuals in the reference group. A new algorithm is proposed that incorporates multiple regression models, each trained on a reference group with a specific age distribution. The algorithm selects the most appropriate model for each case based on a preliminary age estimate of the skeleton; however, this estimate is not used as an explicit predictor in the final model. The proposed method shows relatively high accuracy compared to both the phase-based Meindl-Lovejoy approach and several modern techniques. Estimates generated by the algorithm from primary data collected by different researchers also show a good level of interobserver agreement.

Keywords: age estimation, paleodemography, suture obliteration, regression analysis, reference group.

 

Funding. The study was carried out with the financial support of the Russian Science Foundation grant, project No. 24-28-01050 “In search of effective methods for sex and age estimation from the skull”.

 

This work is licensed under a Creative Commons Attribution 4.0 License.

Accepted: 21.11.2024

Article is published: 15.06.2025