Multivariate Regression Methods for the Analysis of Stature

In forensic anthropology the size of “organs” (bones) from the skeletal system is often used to estimate an individual’s stature via linear regression of stature onto one or more bones. This type of analysis is the reverse of how one usually studies size and shape in biology, where organ sizes are viewed as being dependent on body size. We show that the multivariate regression of bone sizes onto body size (stature) provides a very general framework for both the estimation of stature and for the presentation of evidence on a putative identification. We further show that using an informative prior for stature from the reference sample is completely equivalent to performing the traditional regression of stature onto long bones, that using a uniform prior for stature gives the maximum likelihood estimator for stature, and that using an informative prior stature distribution from a relevant population provides a “population specific” method.