The estimation of human adult age at death is a procedure encountered by anthropologists, archaeologists and forensic scientists. In this research we are concerned with the development of procedures for the estimation of adult human age using easily measured quantities, such as structural changes in teeth. A quantitative procedure for the use of observed macrostructural changes in teeth to estimate the age at death was first proposed by Gustafson (1950). That procedure was based on six age-related changes, each assigned a score on an ascending scale of 0 to 3 depending on the severity of the change. Gustafson's method then used linear regression of age on total score.
Until recently, most attention has been on improving the methods of data collection. However, more recently other methods of analysis and age estimation have been proposed, including the use of inverse estimation as used in calibration experiments. However, as in many applications in forensic and archaeological sciences, the measurements are recorded on an ordinal scale making standard methods of analysis, such as regression, inappropriate.
We have proposed an alternative approach, based on the use of Bayes' theorem which is not reliant on such tenuous assumptions. The method uses information from the teeth of individuals of known age to estimate various probabilities. It is, however, only possible to obtain relatively small collections of teeth of known age meaning that naive approaches, such as histograms, produce unreliable estimates of density functions and resulting unreliable age estimates and confidence intervals.
A refinment to the Bayesian approach uses kernel density estimation to introduce smoothing into the estimation of the various probability density functions. The aim being to reduce spurious results by making better use of the available data.
This work was in collaboration with Dr Dave Lucy and Professor Mark Pollard of The Department of Archaeological Sciences, University of Bradford.
A related project, supported by an NERC research grant, is looking to exploit the same statistical methodology in the area of climate reconstruction using proxy data. This work is being carried out by Dave Lucy (Research Fellow) and is joint with Professor Mark Pollard of The Department of Archaeological Sciences, University of Bradford.
Aykroyd, R. G., Lucy, D, Pollard, A. M. and Roberts, C. A. (1999). Nasty, brutish, but not necessarily short: A reconsideration of the statistical methods used to calculate age at death from adult human skeletal and dental age indicators. American Antiquity, 64: 55-70.
Aykroyd, R. G., Lucy, D., Pollard, A. M. and Solheim, T. (1997). Regression analysis and age estimation. American Journal of Physical Anthropology, 104, 259-265.
Aykroyd, R. G., Lucy, D. and Pollard, A. M. (1996). Statistical methods for the estimation of human age at death. Research Report No. STAT-96/08 (369K) .
Lucy, D. (1996). Human age estimation from skeletal and dental evidence. PhD Thesis, Department of Archaeological Sciences, University of Bradford.
Lucy, D., Aykroyd, R. G., Pollard, A. M., and Solheim, T. (1996). A Bayesian approach to adult human age estimation by Johanson's method. Journal of Forensic Sciences, 41, 189-194.
Aykroyd, R. G. (1995). Calculation of Gustafson error of estimation. Journal of Forensic Sciences, 40, 925.
Lucy, D. and Pollard, A. M. (1995). Further comments on the estimation of error associated with the Gustafson dental age estimation method. Journal of Forensic Sciences, 40, 222-227.