Accurate estimation of human adult age has always been a problem for anthropologists, archaeologists and forensic scientists. The main factor contributing to the difficulties is the high variability of physiological age indicators. However, confounding this variability in many age estimation applications is a systematic tendency for age estimates, regardless of physiological indicator employed, to assign ages which are too high for young individuals, and too low for older individuals. This paper shows that at least part of this error is the inevitable consequence of the statistical procedures used to extract an estimate of age from age indicators, and that the magnitude of the error is inversely related to how well an age indicator is correlated with age. The use of classical calibration over inverse calibration is recommended for age estimation.

Much of the data which appears in the forensic and archaeological literature is ordinal or categorical. This is particularly true of the age related indicators presented by Gustafson [1] in his method of human adult age estimation using the structural changes in human teeth. This technique is still being modified and elaborated. However, the statistical methods of regression analysis employed by Gustafson and others are not particularly appropriate to this type of data, but are still employed because alternatives have not yet been explored. This paper presents a novel approach based upon the application of Bayes' theorem to ordinal and categorical data, which overcomes many of the problems associated with regression analysis.

It is generally assumed that life expectancy in antiquity was considerably shorter than it is now. In the limited number of cases where skeletal or dental age-at-death estimates have been made on adults for whom there are other reliable indications of age, there appears to be a clear systematic trend towards overestimating the age of young adults, and underestimating that of older individuals. We show that this might be a result of the use of regression-based techniques of analysis for converting age indicators into estimated ages. Whilst acknowledging the limitations of most age-at-death indicators in the higher age categories, we show that a Bayesian approach to converting age indicators into estimated age can reduce this trend of underestimation at the older end. We also show that such a Bayesian approach can always do better than regression-based methods in terms of giving a smaller average difference between predicted age and known age, and a smaller average 95-percent confidence interval width of the estimate. Given these observations, we suggest that Bayesian approaches to converting age indicators into age estimates deserve further investigation. In view of the generality and flexibility of the approach, we also suggest that similar algorithms may have a much wider application.

A method is proposed for the calibration of a continuous random variable when the dependent variables are a combination of continuous and categorical, and the model between the controlling variables and calibrated variable is empirically derived. The various probability distributions are estimated from training data by using kernel density procedures with bi-variate normal kernels for continuous variables and uniform smoothing for discrete variables. Bayes's theorem is then used to produce the posterior distribution from which point estimates and estimates of confidence may be made. Individual posterior densities allow each case to be considered separately and cases with conflicting evidence can easily be identified for further investigation. This approach is illustrated by using part of a data set of human adult teeth from individuals of known age. Estimates from the method proposed show less bias than those from the widely used multiple regression. This allows a more accurate reconstruction of the age distributions of ancient populations. In particular bias reduction is most notable at the extreme ages, which also tend to be the least frequent, thereby widening the age distribution. This will allow a more reliable consideration of archaeological and anthropological questions relating to, for example, the maximum lifespan, age-related social structure and the development of age-related disease. Copyright 2002 The Royal Statistical Society.

The recently introduced kernelized expectation maximization (KEM) method has shown promise across varied applications. These studies have demonstrated the benefits and drawbacks of the technique when the kernel matrix is estimated from separate anatomical information, for example from magnetic resonance (MR), or from a preliminary PET reconstruction. The contribution of this work is to propose and investigate a list-mode-hybrid KEM (LM-HKEM) reconstruction algorithm with the aim of maintaining the benefits of the anatomically-guided methods and overcome their limitations by incorporating synergistic information iteratively. The HKEM is designed to reduce negative bias associated with low-counts, the problem of PET unique feature suppression reported in the previously mentioned studies using only the MR-based kernel, and to improve contrast of lesions at different count levels. The proposed algorithm is validated using a simulation study, a phantom dataset and two clinical datasets. For each of the real datasets high and low count-levels were investigated. The reconstructed images are assessed and compared with different LM algorithms implemented in STIR. The findings obtained using simulated and real datasets show that anatomically-guided techniques provide reduced partial volume effect and higher contrast compared to standard techniques, and HKEM provides even higher contrast and reduced bias in almost all the cases. This work, therefore argues that using synergistic information, via the kernel method, increases the accuracy of the PET clinical diagnostic examination. The promising quantitative features of the HKEM method give the opportunity to explore many possible clinical applications, such as cancer and inflammation.