New Approaches to Age Estimation Using Palatal Suture Fusion

Palatal suture fusion has seen little testing and/or application due in part to an inadequately described relationship to formal age ranges. This study presents a modified scoring method that examines fusion on samples of modern documented adult males and compares two Bayesian approaches to age estimation. In the first analysis, American and Portuguese collections were used to derive univariate and multivariate transition analysis (TA) parameters, which combined with an informative prior, estimated age in a sample of modern Americans and Portuguese. For the second analysis, a Bayesian multiple linear regression (BMLR) used indicator statuses as the independent variables with age as the dependent variable. Highest posterior density regions (HPDRs) and highest posterior density intervals (HDI) were calculated for a holdout sample. Final age estimates for the methods were tested for accuracy using cumulative binomial tests at 75% and 90% coverages. The HPDRs from multivariate TA captured age better for younger individuals, but consistently underaged. The cumulative binomial tests on the BMLR results indicated the prediction intervals performed as expected, and we show they are narrower (more precise) and/or more accurate than the corresponding HPDR. The modified method presented here formally links palatal suture obliteration to age using two different approaches, one of which (BMLR) is new to the aging literature. The BMLR provided results free from bias and more reasonable age ranges while maintaining accuracy. We present a look-up table and a free, simple R file for users to download and run their own estimates with BMLR.