Estimating adult skeletal age from a large array of new characteristics and improvements in computer-based transition analysis
This three-year project funded by the National Institute of Justice, USA (NIJ Grant 2014-DN-BX-K007), started in May 2015. It aims to improve Transition Analysis, a method for estimating age from adult skeletons. Precise age estimates are a necessity in biological and forensic anthropology. Transition Analysis has been proven to reliably estimate age throughout much of the lifespan, but is far from ideal. It lacks accuracy, precision, and replicability – although to a lesser degree than other commonly applied procedures. Recent work with experience-based age estimates indicates that it is better to look at a much wider array of skeletal features instead of analyzing only a few structures of high importance. Large set of traits, distributed over the whole body, individually contribute little information yet collectively yield accurate overall estimates.
To define new traits and record their time of transition we examined ca. 1650 skeletons from four skeletal collections in the Unites States, Europe, South Africa, and Thailand. Besides the staff from ADBOU, Prof. Jesper Boldsen, PhD student Peter Tarp, and Postdoctoral researcher Dr. Svenja Weise, the research team consists of Prof. George R. Milner and PhD student Sara Getz from Penn State University, and Prof. Stephen Ousley, Mercyhurst University, all Pennsylvania, USA. Since 2015 the team has collected data in the William Bass Donated Skeletal Collection, University of Tennessee in Knoxville, the Department of Anatomy, Pretoria University in South Africa, the Forensic Osteology Research Center at Chiang Mai University Medical School in Thailand, and in the Bocage Museum (National Museum of Natural History), Lisbon, Portugal.
For every individual a suite of 80 skeletal traits was recorded and their age-of-transition curves will now be estimated by logistic regression. Later in the project this large group of traits will be reduced to a smaller set of easily identifiable indicators displaying steep transition curves. A single composite age interval for every analyzed individual will be empirically determined from the separate transitions. The output will be a point estimate, derived by Maximum Likelihood Estimation, and a 95% prediction interval. Users will be able to choose from different prior distributions derived from national or historical data. The last step is a refined Transition Analysis computer program accompanied by an illustrated scoring manual which can easily be used by the forensic and anthropological community. This program will operate independently on different platforms (Windows, Mac, and Android), but also be integrated in Fordisc.