A Fully Bayesian Approach to Adult Skeletal Age Estimation: Multivariate Latent Trait Modeling With Markov Chain...

Area of Science:

• Bioarchaeology
• Forensic Anthropology
• Computational Biology

Background:

• Accurate age-at-death estimation is crucial in osteological and bioarchaeological research.
• Traditional methods often rely on predefined population priors and reference data.
• Integrating multiple age markers presents analytical challenges.

Purpose of the Study:

• To investigate the efficacy of Bayesian transition analysis for improving age-at-death estimation in osteological datasets.
• To adapt and apply Bayesian methods using NIMBLE and JAGS for skeletal age assessment.
• To explore the method's applicability across diverse skeletal collections and age indicators.

Main Methods:

• Developed a Bayesian probit regression model incorporating multiple latent traits.
• Estimated age-at-death and Gompertz distribution parameters directly, bypassing population priors.
• Utilized transition analysis with software like NIMBLE and JAGS on three distinct skeletal datasets.

Main Results:

• The Bayesian model demonstrated robust performance in estimating age-at-death and mortality patterns.
• Effective estimation was achieved even with smaller sample sizes and fewer age markers.
• The model produced reliable estimates despite missing data, with calibration improving initial age range underestimation.

Conclusions:

• Bayesian transition analysis provides a reliable and flexible method for age estimation in bioarchaeology.
• The approach enhances accuracy across diverse populations and varying data quality of morphological age indicators.
• Population compatibility is not required, only observer-consistent, monotonically changing age markers.